The VFX Artists Podcast
The VFX Artists Podcast
Powering Invisible VFX, with Paul Miller and Marco Paolini, creators of Boris Silhouette | TVAP EP44
This week it's been a real pleasure to meet Paul Miller and Marco Paolini - Software Developer and Product Manager behind the Oscar winning software, Boris Silhouette.
Despite advances in AI, manual rotoscoping and invisible paintouts are still an essential bullwark of the VFX industry and not going away anytime soon.
We begin our discussion with their work on Elastic Reality (used on classic shows like Quantum Leap) and talk about how they have built Silhouette for artists who need to paint 100 stokes a second and export mattes accurate to better than a 16th of a pixel.
Recently Silhouette has expanded it's compositing toolset so we discuss their choices around this. Of course machine learning is something cannot avoid discussing, so we look at both its promise and its limitations at the high end.
In closing both Paul and Marco give some excellent career advice for anyone entering or advancing in the VFX and Software Development industries.
Enjoy!
00:00:00 : Preview
00:01:39 : Introductions
00:02:20 : Founding the team
00:03:55 : Elastic Reality
00:06:51 : Original plan for Silhouette
00:07:43 : The Node Graph
00:10:56 : Marco's Journey from Artist to Software
00:14:01 : The Smallness of the Industry
00:20:05 : What Keeps the Team Busy
00:25:34 : The Scripting Engine
00:28:38 : The Decision to Add More Comp Features
00:31:28 : Performance and Paul's Nightmares
00:35:49 : Silhouette as OFX Plugin
00:41:15 : Limits of Machine Learning
01:06:26 : Retimes
01:17:38 : Paul's Advice for Aspiring Developers
01:20:50 : Learn Python!
01:22:41 : Marco's Advice for Getting into VFX and into Software
01:28:41 : Closing thoughts
(chapters powered by [Chaptered.app](https://Chaptered.app))
#Silhouette #Oscars #scitech #VFX #Podcast #rotoscoping #invisiblevfx
As always, if you liked this episode, please like, comment, subscribe and share!
Listen to all episodes on our website
https://www.thevfxartistspodcast.com/
Thank you for your support!
Paul Thomas Miller and Marco polini the academy presents you both a technical Achievement Award for silhouette I've seen paint artists that that paint 100 Strokes a second it's just just instantaneously there's no delay after a stroke it's like it's like literally airbrushing right um there cannot be a delay between those Strokes I think initially we you know we have the original document from like 2003 where it was specked out as Roto paint and compositing so it was kind of funny because we needed to pull that document out at a certain point when we wanted to expand into compositing and uh it's kind of fun to see now how we had that plan way back when and you know just took a little time to to get there every time I we think about adding a new paint feature that that's kind of deep it's exciting but it's also a little scary just to go go in there and is this going to slow it down that's always my biggest nightmare doing shots doing visual effects you're gonna redo things over and over and over and even doing software you know we're constantly changing it's you can't get upset about that kind of thing yeah we had a facility who was oh I'm not saying I'm not going to say complaining but they were concerned that the rendering in Nuke was a sixteenth of a pixel off when it wasn't silhouette and that mattered[Music] uh so welcome to another episode of the VFX artists podcast we have Paul Miller engineer at bosefix and Marco polini also Bose fex both working on Silhouette so we're going to talk about the history of the software which is vitally important obviously as part of the Roto prep pipeline so um do you want to introduce yourselves I'm uh Marco polini I'm the product designer and product manager of silhouette cool Paul Miller a lead engineer on Silhouette since we uh released the first version in 2004 I think are you both have you both been with the product from the beginning or did you join some of you join later we've we've been uh we're like the original team yeah I think it was back in 2004 we were had a sep other company digital uh film tools and at the time uh I was also part of digital film works and Paul was part of another company profound effects and uh we got together uh because uh you know in in the course of our work uh digital film Works uh whereas a visual effects artist at the time you know there there was really kind of a a need in the market for a good Roto paint program and the Roto paint programs that had been around were no longer really around and supported and uh we just felt like it was a good time to to resurrect it absolutely so um will you did it start as an in-house product or are you immediately going as a commercial product to sell to vendors it was It was a commercial product uh from the start and uh what what made the product was you know Paul's expertise with uh elastic reality and really had a great uh product there that did morphing and was also used for Roto as well um you know we were able to build on that expertise from his you know technical expertise and then on our side we you know we kind of have the artistic uh you know visual effects background and I think that the match was really really good you know on on top of over the years all the input that we've gotten from artists around the world really cool so yeah elastic reality um I mean this has actually come up a lot of in our episodes where we've gone back in time with uh I went to Sean Cunningham as well talking to early basic feature domain and obviously fighting important and because all those big morph effects were pretty much done in realistic reality I think even that Michael Jackson video you know there was there was pdiar damages they had just that was their in-house tool and they that was about a year before elastic reality came out so we so elastic reality you work there as an employee yes a company at the time was called asdg uh that we came out with elastic reality in 1993. uh that was my first software that I wrote for silicon Graphics uh workstations uh and then the following because it was so successful the following year we renamed the company to elastic reality and then a year after that we were bought by Avid technology and so what let's spark the idea of you know going off on your own starting your own business is obviously something a lot of people yeah I would think about and most don't yeah I think I think Morocco I know it's interesting we're both at Avid around the same time this has been about 95 96. I think Market was the first one to leave I think right you went off to uh the post group uh you know it went off to digital film works and uh there's a little filmworks yeah and then Paul you went off to uh profound effects after that right yeah yeah I worked on a last year holiday for a couple more years at Avid uh was the lead engineer on their titler uh the 3D titler project there that was called Marquis um then I left we've moved back to Wisconsin uh joined forces with my old boss from elastic reality and we created profound effects making plugins for After Effects and avid um so while at Avid I designed their plug-in architecture AVX so there was a big demand in the market it was the first plug-in architecture for the media composer editing system um so uh Marco actually reached out to me when they were at digital film works because uh they were developing their own in-house tools for compositing and keying and things like that and they wanted to turn them into products that they could sell and so since I had done the login architecture and had just started my own company doing plugins uh if that's how we reconnect it as well I started doing plugins for Marco and to the film tools so my own plugins for profound effects and then eventually we just decided that maybe we should write a rotoscoping and then and a paint system at the same time it was always planned to do I think the original idea was to do a compositor um there weren't that many around it was back in the days of shake and very early nuke uh so we thought maybe we'd do a compository but we'd start with that that bread and butter the thing that everybody needed was rotoscopic I think initially we you know have the original document from like 2003 where it was specced out as Roto paint and compositing so it was kind of funny because we needed to pull that document out at a certain point when we wanted to expand into compositing and uh it's kind of fun to see now how we had that plan way back when and you know just took a little time to to get there but I think the fact that we did plugins first allowed us to kind of build all the little bits and Bobs that were needed and I've cut of course Roto and paint were you know for us the kind of the main components and we already had all the other plugins and so I think Paul wasn't silhouette from the very beginning even though it was just rhodo and paint was the architecture was no base wasn't it yeah we designed it internally as a node-based system uh keeping in mind that we had all these plugins being worked on at digital film tools and um we knew that we were going to want to add more compositing features so internally it started out and it has been all the whole time a node-based system and um at first we had a limited set of nodes the first one was just Roto and then the following year we added paint and then there was uh I think Ikea and eventually morph um and then version six uh we exposed all the rest of the nodes we added a whole bunch of more nodes out and just exposed um the tree and and everything else right so the nodes were more like um I guess a hypograph in my sort of hidden away rather than someone like nuclear Houdini where the notebook is like front and center of your winter they were they were they were kind of there where it uh you you couldn't rearrange them it was kind of a set order where it was like the source then Roto then I think paint and then we had like an effects node um that was kind of a combination of rhodo and paint and and then morphing um and then you can kind of you could turn one node on or off or you could have them all on or you could view one and edit another but there was no flexibility in how you place them you couldn't have multiple rotos you just one of each yeah it was just a chain um but internally it was literally creating and destroying and reconnecting nodes it's just you couldn't you couldn't directly manipulate them yeah it's interesting visible notes I never really I thought of notes purely as a UI thing I never really thought that nodes could kind of exist in the background of a software that isn't uh node-based as a user yeah I think I think once once we decided to do the compositor Paul wrote the the actual Tree View or what we called nodes uh separately so there was a separate little program for you know connecting up you know the foreground the background that you know the obey mat input uh you know so that was that was all done first um and I think believe I believe and that was yeah yeah that was mostly just the same time uh so we could experiment with how we wanted the node interactions to work so that you could you know how you would connect things and getting the behavior right where you can I don't know if a lot of people know this but you can you can like we we call it well I call it internally um kiss detection so if you put a node near another node it'll it'll connect up to it if you get it close enough um little things like that but that was mostly just to save time avoiding the compiling and running cycle you know because so that was was pretty big by then uh we got all that sorted out working and then integrated it and that was that was really neat when we started where you could actually create more than one note that always worked internally you could have two or three rotos there's just no way to to get at them yeah uh so that was pretty cool Marcus and Demarco is you came from a narcissist background I was just going to ask how you ended up at Avid and and then obviously starting a business as a software creator rather than as a supervisor or something yeah I actually worked in Hollywood as an editor first as a just a regular online editor uh using you know traditional linear editing systems and um the facility that I was at ended up getting a media composer uh as well as a Parallax Matador paint system for doing Graphics at the time and I uh you know as part of the package I got training for Avid uh and when I went to the class uh the instructor said hey you know because I actually went the class I went to was like a train the trainer course where they were training instructors and um they said you know if you ever wanted to do this if you ever wanted to train people you know we we have a need for that um and so at a certain point I you know I I I did it um kind of on a freelance basis going out to various uh facilities training and then they offered me a job and I took it and um you know I did that for a couple years and I was a trainer to start with which also then led into product design because I I knew the The Matador paint system it was actually a user uh and so I was involved with product design uh for a brief period before I left Avid on the uh illusion compositing system we could call that Avid media illusion oh wow you know they had it for a you know a couple years you know they they were trying to kind of compete with flame at the time you know where flame was really fast uh you know where there was you know quick interaction where you'd be working with the client over your shoulder uh they were trying to compete in that area eventually the you know the product was discontinued but uh from there I ended up uh working with one of the Avid customers which was digital film Works who had uh uh Matador they had illusion they also had Shake um and that was my kind of introduction to feature film effects but I had you know after working at digital film works for uh I don't know maybe four years we had a number of tools that we said you know what if we could get a programmer and start working with them we have some tools that you know other people would like to use and uh you know that's when we hooked up with Paul and started getting the plugins going we came out with a our first set which was composite Suite we had various composite related plugins and I I think we came out I want to say we came out with the Avid version first and then after effects yeah and the the light filter the light effect was the very first one that we did yeah I mean we because you wanted to emulate you had optical filters uh and and so it was somewhat cinematography based where you know Paul's speaking of the light filter which you know gobos and it was quite an extensive filter from which a number of other filters were created it's pretty yeah sorry to jump in I just I just wanted to point out um so digital film Works um the owner one of the owners of that company was also one of the sort of like key customers of elastic reality back in the day so it's just to show you how kind of incestuous the whole industry is um yeah so in a sense that I've been working with uh the um the one of the founders of the film Works uh Peter Moyer says like 1993. uh he was always providing input on elastic reality and uh other tools that in the last reality that we had we had built a lot of them used on early television digital effects just wanted to kind of shout out for you I just wanted to sort of I just wanted to help um some of the sort of younger listeners or viewers just because a lot of things come up that might not not really covered in the industry even if you're a work professional so obviously linear uh editing is basically you've got tapes you've got a tape deck on one side and you've got a tech deck on the other side that records and if you want to do a resolve you have two tape decks on the input side and like a mixer like a DJ would have to train them and you would record and you'd do it in order um in a linear way which was used for episodic TV and the non-linear system at the firm's first film when you're physically cutting the film and then the digital tools like the Avid um and then I guess the other thing that came up was the software's Matador so it was one of the earliest I think there was there were like two or three paint tools at the time like math tools Matador is one of them and I can't remember for the life of me remember what the other there's uh you know there was in-house Liberty yeah like Chiron had that at a certain point right yeah Liberty paying um no but it's interesting too that Paul you mentioned you know that the people that you work with and somewhat incestuous but my my first job when I was linear editing first I was tape operator where I was just spooling up the tapes in the machine room and you know getting food for uh you know the editors and the clients but you know I got my chance offline editing first uh one of the other bays and I I worked with um price pethel who ended up being doing you know effects of digital domain um on a number of films uh but you know working with somebody like that uh worked on a Diana Ross music video who you know and he he was obviously a number of years ahead of me but being able to learn um from you know guys that had had it all figured out doing effects in a super rudimentary way in a linear uh you know edit Bay where you know cutting mats were you know using a down shooter camera with a black and white you know something that was drawn or printed out to be able to cut the mats but um and Peter Moyer who uh who was a partner in digital film tools and uh was one of the owners of digital film works he did all the effects for uh you know the original Star Trek um generation uh did a lot of the Michael Jackson videos uh and he was the head of the post group uh the digital division uh where you know Rob Legato was working at the time uh you know just a number of people you know and then they all split off and do uh various things so yeah I think we've all been pretty lucky to kind of be in the right place at the right time and you know having met certain people and just some some is just luck you decided to take a job and you met incredible people that helped you along in your career yeah no it's really important I mean I think it's the biggest thing is is the relationships you're building and it is a small industry even now I mean it was much smaller then and it's a lot bigger now but it's not that big like you will especially as you get more senior you will meet the same people throughout your career it's undoubted like the other people that are Juniors with you right now if you're a junior or mids review of your mid or even senior of your senior they're going to be around and you're going to move to another company and they're still going to be around you go to another country and there's still going to be another round so you need to like build those relationships because as you you know as you yourselves have done it can it can be vitally important yeah I'd like to real quick I'd like to point out something you made the analogy you're talking about uh non-linear and linear editing there's a there's a parallel too right with with filmmaking right you went we went from Optical effects to digital effects and the same process goes with the optical effects right you feed the film through put an effect on it or you might key another element over and then you just do it again you run through a strip of film and and that's how it was all done until late 80s right early very early 90s and then things started going digital um and you know it's kind of just kind of well it's actually really fascinating and pretty awesome just to be part of that that switch over when things started there were things were being invented uh all the time for how to try to do this stuff and like what I when I started working on elastic reality in 1993 I didn't even I didn't know what an alpha channel was um I hadn't met those people yet um you know it was an established Thing by then but it was kind of interesting to be like what's this Alpha Channel thing and then and then you know a year later you've got this software that everyone's using it's um pretty neat to have having lived through that but on the flip side like I've been doing Roto now for almost 30 years it's almost kind of sad in a way it's um but it's interesting that I mean but Voda is so fundamental and I I guess we're on the verge of something similar right a similar paradigm shift right now and I know that we had trouble scheduling this interview because you guys were busy so I was wondering what what you've been up to we do so many things uh you know we just point releases yeah Point releases we're working on a solo at 20 22.5 uh release which will come out shortly and then we're off to our 2023 releases but um you know we're constantly taking user input trying to improve the product um we you know it's always like a balance of like you know what we think we need to put into the product to keep it viable to make it effective make it better and and then what users are asking for as well you know because we we have both large companies uh that may have different needs than say you know the one-man shop so you know we're always trying to weigh those um but you know we're working uh individually you know on our product and of course Boris effects is uh working on you know as as many image processing type projects that you could think of whether it's AI based whether it's just traditional you know grain management uh there's you know research and development happening in in tons of different areas because we you know we have not only their silhouette there's mocha Pro which also does Roto and of course their tracker but their Sapphire plugins the Continuum plug-ins and uh you know we always try and think of things that can be shared company-wide between products and and also now being part of a larger company that has these different products we as as silhouette as a product benefit from that as well so if uh one team like the Continuum team who now uses kind of our silhouette architecture behind the scenes that Paul wrote if they make something new we automatically get it if we want it it just can show up natively um and and Beauty studio is an example of that for doing you know Edge preserving blurs and Beauty work on people's faces um you know to lighting plug-ins everything else and that goes for Sapphire too because we we include the sapphire plugins as well well so how big is your team your uh it's still predominantly me um and Marco um we have some uh some some contributors that are working on smaller pieces like importer exporter type stuff that's stuff that's shared around the company but so far the core like pretty much what what you see in silhouette is is uh is my code uh except like like Marco said in for 2022 we added a whole bunch of stuff from like Sapphire and we had a particle illusion um and those really expanded our you know our tool set quite a bit um in terms of just nodes there's a lot more effects nodes and more compositing nodes and things like that so Paul what's the uh you know we always talk about you know whether or not if you had four more programmers or five more programmers you don't necessarily get five times more output there's a certain uh benefit of one person controlling everything there is there is every we've looked at expanding the team for a long time and every time it just seems like a daunting task to try to train somebody up on 15 year old code base and things like that but there are there are smaller more low-hanging fruit uh type things that that that would be pretty straightforward the silhouette's pretty modular from an architecture point of view and I designed it so that people can plug into it easily and it can be expanded easily and and nodes can be built in three different ways uh you can write python uh to make it do a whole lot of stuff um so there's all that but lately we've going back to the feature set and how we talk to customers we it's it's really cool because in the four in our forums you might get a couple of can I do this kind of things that are really easy that you know I'll just be in the middle of a big feature and then like wow yeah I'll put that in and then the next maintenance release comes out it's just there and then we've got some other uh you know bigger facilities that want like bigger features that can take weeks or you know months or whatever um so it's a nice trade-off in terms of development time but yeah we would like to definitely expand the team uh Silhouettes doing well and we have like years worth of features just in the list part of the thing part of our problem is deciding what's going to go into what release um you know we originally I think we were planning on putting a bunch of this stuff in this 2022.5 release into the 2023 but we like we thought well we've got so much stuff let's just put it into 2022.5 you know mid mid-year release um we have so much stuff planned for 2023 like let's just split it up and you know it's already done so let's kick it in there and it it's it's cool it's it's fun and I guess having it earlier I mean it obviously adds value to the customer but it also gives them a chance to feedback on it if they want yes what what once it gets out to the artists it does change um I guess from what you imagine people always do things you don't expect them to do with your software yes exactly that's one of the reasons for instance why the key binding I always talk about this like the keyboarding system is a great decision way back in the day to just build it on top of the scripting engine so we get a lot of requests for can I make the software work like other software that I've used and the answer is yes there's no easy way to do it in the GUI you can't just go in and like grab a button and assign it to but you can go into the scripting system and literally change the way every single button works and what it does um and we've that's really worked out pretty well for a lot of the bigger facilities who that have integrated silhouette completely into their pipeline when they you know they get a shot assigned to them from some database and it's all set up and everything ready to go they don't even have to say you know import media that's really amazing um and we only barely know the the surf the scratch the surface is what some of the companies are doing well and it's not it's not just assigning like a simple feature to a keyboard shortcut there's uh you know the concept of actions which are scripts we can do very complex steps multiple multiple things in a row um you know and you know Paul's got a set of actions that come with silhouette you know that kind of serve it as examples uh but you know very complex things can be done some facilities have done their own rendering system that pops up their own user interface that completely bypasses what silhouette does um yeah renders right to a database or you know asset management system for instance um uh lately a lot even more as Marco said a lot of even the features of silhouette itself like even a whole bunch of the road Roto uh tools that we added in 2022 are written in Python that just show up in menus but you can go in there and change the code or see how it works add your own stuff it shows up on the same place that's been really fun I'd like to make it as extensible as possible so I my question is you see now you uh are talking about finally getting to do some of the compositing stuff um I think it's really interesting because obviously your big customers they I mean silhouette is pretty dominant in Roto departments and to some extent impact departments and apart from where and so where is compositing obviously new keys is the big sort of ruling the roost and the one-man chops as you say often After Effects so you know I mean I can sort of see that you've got this advantage of have this advantage of not being a competitor to nuke in just doing Roto but there's also you something quite exciting and yeah something I'm interested to see another what another compositor would do and what might be the advantages and and how to you know initially initially we had a we had requests from various you know prep departments at larger facilities that were doing paint Roto uh a little bit of compositing and wanted to be able to do the entire shot in silhouette and not have to go to Nuke um you know there's there's obviously different factors some of it's easy to use some of it's training some of its cost um you know for the seat they don't want to tie up a nuke license to do certain things or a particular artist really only knows silhouette so we we initially excited to do the compositor because of those requests where like if we could just do some basic compositing we can do a little this little of that uh we'd be happy so initially that's why we did it and of course we've added more and more to it and the more we put in it the more we get asked to add to it um because now the requests are like if you could do a little more of this 3D stuff that'd be even better um you know so it it you know once well now that we've gotten to a certain point then we get the additional request but I think that's that's why we originally did it we wanted to try and keep those departments and customers in silhouette if possible on those particular shows um uh are you talking about you know uh removals paint fixes uh you know procedural uh removes where you wouldn't want to have to paint every single frame maybe you paint a Clean Plate and then you use tracking and compositing to you know comp a whole block um you know it's a lot of it it first came from the paint department saying because initially they were removing grain and they wanted to comp the grain after and uh you know paint on the degrained plate and you know then it was like oh we need multiple Roto nodes and and maybe we're going to use you know a couple transforms to track in a couple major spots but then paint a little bit more here so it was a more involved paint fix and you know whatever the paint fix may be whether it's you know removing a wire removing um you know fixing up you know a head replacement or what a whatever you know it um we're seeing a lot of tattoo Replacements yeah blood Replacements things like that yeah I mean those can get pretty calm heavy I'm I'm going to be honest I've not actually used to it I've always used nuke from early on and at the beginning I shake uh photo which wasn't fun I mean it and that obviously depends on the kind of stuff you start out with I think silhouette is biggest in the big facilities yeah it's just extensively for paint yeah for paint work just because the paint is so fat efficient and and it's more of it's much in performance wise you've got the advantage of and you when you are doing rotor and paint like you you're just simply playing it back faster yeah like the first for instance the Roto if you when you have lots and lots of shapes and you're using motion blur I mean the differences between multiple seconds in Nuke and milliseconds in uh Roto for changing frames I mean it's it that doesn't even compare on on the pain side it really kind of depends on maybe how many strokes uh you know Vector systems when you start to get a lot they can start to lag um you know so we're we're not Vector we're raster but we we have some changeability and uh reproducibility and repurposing of Strokes that can be replicated over multiple frames um you know I think it's integrated tracking yeah integrated tracking integrated Roto in you know into the paint so you can be obeying the mat of the Roto as you're painting you know that's all integrated together so it works you know very nicely um and of course this all works in Stereo so for the for the shops that do stereo it's really important to have those tools and that those stereo tools came from a facility uh request saying hey we're doing a show we need to be able to paint we need we need to be able to Roto we like to do XYZ and we need to buy this date and uh that's we we ended up doing it and I think at the time I I can't remember Paul if we were the first to have stereo in that area but it was it was pretty close with whenever nuke started doing it I'm not quite sure about the timeline yeah I don't even know what nuke's workflow is for stereo Roto but um I thought we might have been the first where you could you could Roto either left or right eye duplicate it to the other eye and then simultaneously edit both splines on both eyes if if you want on a per Point basis and then follow the transform paint similarly it was the same thing yeah paint you can paint on the left side the right side both sides at the same time you could transfer your paint from one eye to the other and then it would take into account that the interocular distance and combine that with the match move that was all in version three that was our big 3D stereo uh version wow I mean it was 2D only there's no 3D objects but it was just the ability to stereoscopically do everything no it's just interesting and a bit um obviously I guess stereo is um less common than it was a few years ago but other um has there been any kind of um advantages of building those tools in non-stereo projects well because they came in very handy when um companies started doing the stereo conversion too um no but I think that's not that's not quite the what you're looking for um are there still definitely many shows that are starting shot in Stereo where they're still doing everything in Stereo um one high profile one uh that I that I always think back to was um WETA um what was the they shot it 120 frames a second oh man remember that and stereo no no they did that in 48. this was the uh Ang Lee movie um remember that Marco yeah I don't remember which one it was but yeah they yeah they should they shot that at 120. stereo and they had to do all the they did all the paint Roto work for that with an astronomical amount of work so they were you know they were asking for things to help with that the shots were so long um they we had to make a few tweaks to you know make that manageable if you had to apart from the fact that I think price point wise rotating and you maybe isn't the best thing to do anyway but once you have those very long shots that's where you would really also start to feel the pain of trying to use me well and and now we have uh Silhouettes and ofx plugin so it will run as a plug-in within nuke um of course we can pass our shapes back and forth to new through an export process but now you can launch silhouette from inside new work on it and and then you've got if you want to work in Nuke that way you can still get into silhouette work on the shot come back and then you have the benefit of the solid user interface but the speed of the silhouette rendering and you're ensured that the what you saw in silhouette will match exactly to what um you see in Nuke where you know the shape export process uh although is pretty uh pretty established and it works pretty well there's always that situation where a shape is doing something that it shouldn't be doing in a particular situation be because of the differences between the products um you know whether it's you know the interpolation method is slightly different in in some cases it's you know it's there's a there can always be a case where it may not be right you know somebody will always find that yeah we had a facility who was oh I'm not saying I'm not going to say complaining but they were concerned that the rendering in Nuke was a sixteenth of a pixel off when it wasn't silhouette and that mattered that mattered to them that sixteenth of a pixel and at that point it's just it their renderer is just different so I was amazed that we were actually got that close um but that this is something that the the plug-in completely alleviates if you need that extra precision and it's their performance hit from running it as a plug-in so I guess you've got to wait for nuke to kick the the media to silhouette oh I don't know I haven't seen one I think it's still faster at the end of the day there's minimal conversion going on um uh it it is a standard of effects plugin though it's not it's not a native uh ndk effect so it probably could be made faster but probably most of the overheads are going to be in silhouette itself not that and in the the silhouette plug-in whether it's in After Effects or in Nuke it's it's the full product now so you you have a node-based compositor Roto paint all the associated nodes within those uh within those products so it's it's kind of interesting that you can get node based compositing within After Effects when it's layer based so you can have a completely separate uh node-based comp going on in AE separately I mean that's very powerful for I mean that changes the game with After Effects a lot because obviously paint and rotating After Effects is also a bit more tricky to control just because the nature of it yeah like you know we really did set up the paint system for instance so it is really designed as a classic like hand paint kind of system and you know some of the some of the artists that have done shots as from a handpaint perspective right where they hand paint every frame in the traditional manner has been incredible um you know they'll have thousands of Strokes on every frame um it was designed to handle that and it it does it really well but I I'm continually impressed to see what that you know you don't a a host workflow May might change your behavior you know if you know you're dealing with Vector you may do something different if everything's a vector paint stroke or something or if you know that you can't have more than a you know you get to a certain threshold and it's just going to start slowing down and not be usable um that that matters significantly to you know some of these artists uh and the same is true for the Roto I think it's interesting to see to um you know we've seen some like sped up uh screen captures of an actual very very complicated shot being done in silhouette with the paint and it's amazing to see the process and how I remember when I was painting things you know you'd spend a certain amount of time in each frame then you kind of play it back try and fix problems but these really complicated shots is you know these artists just beat those shots down until they look perfect it you know first you're looking at it going no way it's gonna look like crap and they just keep going back and back and over the frames and until the lighting's exactly right and it's just amazing to see what what can be done in the hands of somebody who's got that skill it's pretty amazing I think very cool and I guess um so far I'm talking about the Paladin a bit like the analog to digital I mean we've got some really impressive machine learning stuff and of course the Roto that you're getting from AI at the moment isn't up to the stands of human voting on the same with the paint cleanup but it is getting better and it's getting better fast so I was wondering what your thoughts on that and what you're doing about it in silhouette I guess I'll I'll start by saying that you know it for us we have made sure that we have what we feel are the best set of manual tools that can handle a person running in the forest with a shaking camera against changing flickering backgrounds and leaves and you know we've tried to build the tools to handle you know the ugliest and most difficult shots as far as roto's concerned so when we think and I think consider AI at least for me um I think there's you know the questions that we have to ask are is a mat only okay for output um if it's not can we create a shape with as few points as possible with this few keyframes as possible so that it can be edited later especially in the areas where it's not right um now for some people they're like I don't care Matt's fine um other companies or people need those shapes um I mean if we assume that it's let's assume it's perfect nothing needs to be done to it well either a mat or the shape regardless of keyframes you know then that would be okay I don't know that will ever get to perfect uh in all shots um but um you know if there's a way to accelerate the process to give you a great starting point that's editable well that's that's a different story um and you know that's where I think for us to be interested that's that's what we need um and you know we've done some things for customers who are using AI to spit out a result but then they need some good manual tools to fix the overall result over the course of a bunch of frames with a ton of keyframes and and uh I think in yeah a ton of points so in 2020 I think it was in 2022 we put in a um a multi-frame uh range feature that allowed you to attenuate the effect of your adjustment over you know a ramped area of keyframes so that if your AI result drifted say consistently over 200 keyframes you could kind of like slide that back in um but yeah we've you know we've looked at it um you know we'd like to have the magic result just like everybody else but but again it's trying to figure out you know how much memory does it use how fast is it um and is it editable to into the result that you need it to be so you know we're we're looking at a lot of those things and uh you know we just need it to be right to be useful because at a certain point if you get a result very quickly and an automated result say in an hour uh but it takes you four hours five hours to fix when you could have just did it in three hours manually um you know it's just it depends on when that starts to change you know that it becomes more important I I think the editable um feature I think that is is that the tough nut to crack because that's what I found with every time I've tried experimented is it either works or it doesn't and when it works great when it doesn't well what do I do I start again I just go back to the way I was doing before yep but what I want is because the starting point from AI that I can work on yeah there's a little idea it also depends on a lot and there's different thresholds for what constitutes quality and and good enough right and as I mentioned there's some places where a sixteenth of a pixel to them is noticeable and if your AI solution isn't you know up to that level where it's perfect then you're you're redoing it anyway um this this makes me this kind of takes me back probably 28 years to elastic reality we added a feature uh where you could where it was the one of the first Edge snapping implementations so rather than dropping down bezier points you could just draw a line around something and then it would snap you know it would do an edge detection while you do it and it would snap but if you remember some some products have that we even have something like that in silhouette um but it always bothered me because it didn't produce it didn't put points where the artist put points it would right over the years like when you're doing Roto you know you kind of kind of develop a sense of where where do you put your points and it's just automatic when you're laying out shapes and things you know where to put the points and then you know there's the temporal aspect of that but a lot of these tools even the the automated um you know the AI to to point uh shape shape conversion type problem is gonna is going to have that issue where it's not creating the points and the shapes where the artist would do it and you'd have to go end up having to go back and tweak them anyway um this has been a pervasive problem for like this whole time where whereas I think it's all about your threshold for what's good enough and one of the things I learned I don't know I think it might have even been Peter Moyer who told me this I think one time I was working on some feature for some product and I said I think this would be good enough and he said if it's if it's just good enough it's not good enough and it always stuck with me so we we're we're making we're making tools for for artists where it's got to be perfect not just good enough and I I think AI is still in that well into that realm of it it might be good enough but depends on what you know you're going after I think some of our customers are doing um previz on kind of quick comps you know prior to a film's financing uh as a as a method to quickly comp um you know tons of shots and having a AI mats that are not great but can be generated somewhat automated in those cases are good enough for them to say okay we know the effects kind of look like crap but it's enough to kind of give you uh financers and Studios enough to decide whether or not you're going to give us money to finance this project so I think you know having a lower Quality Quick result um you know for that type of work is you know great you know just like offline editing where you you know used to work at a much lower resolution it was like good enough to see the cut and good enough to get approval and move forward so I think these these are all tools and you know we we want to embrace them just as much as everybody else but we also you know as Paul said there's a certain level of quality we know our customers want and if if the majority of the time it's not producing a result that they're gonna use then they probably won't use it and like Paul mentioned like the snapping tool that we have in silhouette um that came from mocha um I never use it because of exactly what Paul just mentioned I'd rather I'd rather spend a little bit of time and get those points where I want them and not have a ton of points that are randomly placed that I then have to drag around and put them where I want um so I I think for us you know like in the short term you know of course we want to make sure that our tracking is as good as possible because that kind of assistance uh can produce those you know as having the tracking for information as part of like a transformation layer and the keyframe um for shapes on a separate transformation that allows us to have the ultimate flexibility so I think kind of the shorter term answer is to really get that tracking locked down as much as possible and that's why you know in addition to the point tracker in silhouette we had our own planar tracker we include mocha Pro uh both got kind of an integrated fashion and but also as uh ofx bundled plug-in with silhouette so you've got the best tracking available directly within silhouette so and now we're starting to take advantage of their mesh based tracking which adds that extra nice layer of organic tracking and we're still innovating in terms of um just being able to move points around just um as the number of points have gotten become larger and the number of keyframes have gotten larger and as Marco mentioned the client that uh is producing their own AI generated shapes they need to edit uh in silhouette now they've got a whole bunch of control points on every single frame and how do you edit that and that kind of forced us to think about the problem of how do you and Roto all that stuff all that data and uh added we added a bunch of tools for doing that um came up with a a really neat uh Roto um uh combining the paint brushes so you can do brush shaped brush based reshaping which is really nice and organic and lets you kind of pull things around and you combine that with multi-frame mode and the ability to do the uh the multi-frame ranges as Marco mentioned with the attenuation you can get a lot of really nice oh and then then we added the preview so you can preview motion blur or over time without rendering it um so all of those things combined give you a good sense of what you know what your shapes are doing over time uh and then be able to very quickly edit them organically it's uh it's keeping us on our toes but it gives the artists those extra tools they need you know do those kinds of shots that's really cool I just uh kind of circling back to machine learning just quickly I was wondering is that something that rather than trying to do the road to or do the um do the shots is it something you've thought of using to make some of your tools like the tracking better or the snapping actually I would imagine if you're training it by putting the points in the same place you could train it to have better snapping when you try using snapping we've tossed a lot of those uh things around and um a lot of it comes down to development time and having enough people to do those kinds of things I'm still focusing on stuff that you know a lot of our facilities are actually asking for um I know you know AI Roto is a Hot Topic but at the end of the day we don't get a whole lot of requests for it because they're um you know they're they're already kind of invested in the tools and know how to use them and and again have to get that level of precision we talked about but we say we have a AI uh team that's you know working on various models to assist um both tracking and Roto but you know there's all the things that I mentioned um you know are going into is it usable is it fast enough uh can it be deployed properly um you know we we saw some great tests um earlier um from somebody we were working with and um you know on the face of it it looked it looked good but it was like it it took forever and it used way too much memory and the edges were not good you know but in general in general the motion looked pretty good you know but you know aside from the fact that you know hands disappeared here and there it you know heads you know were halfway it looked great until you until you see zoomed in it's like okay how do we like overcome these things and it's like you know some of the things are oh you gotta go multi-year projects to try and solve so I mean I you know as he mentioned things are changing rapidly it just depends on what's changing and how rapidly you know I guess it's the YouTube creators and that are gonna really lead the charge with this stuff because they can get away with it good enough and possibly yes be at the Forefront and then of course as it gets better they're going to be the ones that have been using it for four or five years when maybe in four or five years it is good enough yeah and as so did that sound like oh for you I mean I imagine that you're going to keep developing your products in a in the market that is there in keeping up to date with obviously you've got your own AI research projects and you're moving into comp and is that yeah we're going to keep continuing to do what our users want us to do for sure um and if more of them you know they're smart they know right they know that they've seen these tools they they know that they're not going to produce the results that they need so but when they do get to that point they'll probably start asking for more of that integration better integration and then we'll then we'll do it I mean but yeah right now we we have these research projects going on and they're ongoing and it just I mean to me it's kind of like having a keyer in your product that just produces a bad result yeah and it's you know so that's hard to get excited about um but if it's a key or that produces a great result then of course you want that and and our customers are focusing on churning out their shows that are going to Netflix that are going to theatrical release and you know in AKA yeah yeah and it's um yeah I think the conver I think the conversation we should have the conversation when at which point when and if um AI produces as good as a result as uh for Roto as a human can on the majority of shots and if we don't have a solution by then then we probably won't be talking to you anyway so what um yeah it'd be nice what but see this this isn't this is important though too to to realize like so like when I started I started linear editing and all those linear editing companies they're gone you know and so the avids and you know the adobe's came on the scene I had a good friend of mine whose family uh did process photography I was the Hansard family who who did all the rear screen projection um in Hollywood for for years and years and years and it was interesting because as I was starting at digital film Works they you know that's when the digi in in the 90s the digital uh really started to pick up and they essentially their business turned into nothing same thing with you know Optical film Labs uh and Optical effects so I think that we need to you know keep an eye on these things because we don't want to be the linear editor uh the process photography guys you know we we want to keep iterating um the product as we go so that it's it's always usable um you know for for whatever is taking place but we also see our customer shots and we're like no way you wouldn't even you wouldn't even attempt to run it into uh just because the amount of motion in the background and you can't tell what's what and the motion blur uh it's like forget it but like you said certain static very slow moving shots um you know that like Adobe may show with some of their Sensei technology you know somebody walking across and doing some color correction yeah that may be great for a lot of customers doing corporate work and they don't need that 16th of a pixel accuracy and they don't care about kind of funky edges every once in a while um so yeah I mean I we I don't I don't think you know we don't want to come off as poo pooing it we know it's there we know it's getting better and uh we we'd like anything that can make our software better and cut down on the amount of work but again has to be good yeah absolutely um and uh I guess the other question is what technologies are what other Technologies are exciting you in your field you know we're doing research uh in Grain management uh because that sounds real exciting it's a big it's a it's a it's a it's a big request see they see some of these things aren't exciting but to our customers they're like this is kind of like it's like how some people feel about Roto it's like it's not exciting but it's one of those things that they have to deal with on a day-to-day basis and it's important for them to be able to extract it properly to add it back in properly uh with the least amount of problems um I don't know I mean I think for exciting stuff I mean we're always looking at you know how how we can propagate our shapes automatically with you know again as few key frames and as few shape points as possible uh and we have a number of you know areas of paint that you know we're investigating that we'd like to do as well yeah there's a lot left in pain um and that's that's a big it's a big project but that would be that's exciting pound paint is hard I mean I think um people outside of the industry or even outside of 2D you know they're like well Voda paint is Junior your Junior development you move into comp but that's not because paint is easier it's because I mean there there's a sense that there's a there's a very definite right and wrong which in comp isn't always true because you can be more subjective but the reality is some very hard paint shots are just very hard and can be very senior you know there are people that are seeing a paint artists um who are given the shot that no one else can do and none of the nothing can track and they have to you have to paint it frame by frame so yeah you've just described our friend Eddie at WETA he does some amazing stuff um and he's whether one of the the best artist requesters of of things over the years um it's always been very valuable ever since I was elastic reality having the Evans certain key users that are kind of vocal they're not they're not mad or they don't get upset with you but they're just like hey can you do this can you do that you kind of want to you really want to please them uh and but yeah certainly and you know not even that I mean just from the from the Arts perspective it's just thinking about some of the math required to make some of the paint work this kind of keeps me up at night uh when you're getting into like like the paint engine and silhouette I I I'd forgotten about this part of it but you know when you're with because of the way match moving Works um when you're painting you're not always painting on a flat surface right when you project the paint onto another uh frame of motion your brushes start to become uh oblong ellipsoids and in space right or every sample of a brush starts to become non-rectangular region in it and uh the maps into all the other math they're like cloning and things like that um it's it's complicated and every time I we think about adding a new paint feature that that's sort of deep it's exciting but it's also a little scary just to go go in there and is this going to slow it down that's always my biggest nightmare is this going to make it slower because you need to if you're painting frame by frame you need it fast because you need to be back and forth back and forth otherwise it just boils if you try painting fan by foaming nuke and you don't really have that playback possibility um you know that it's just going to boil when you play it back um yeah and not even I mean you can try it procedurally you're trying to do something else um anything but paint if you can yes yeah in a silhouette it's getting baked in right it's getting baked in it it knows all the data and you can use that to your to whatever you want to for later but it's getting baked in so you don't have that problem but you have the problem of when I I've seen paint artists that that paint 100 Strokes a second it's just just instantaneously there's no delay after a stroke it's like it's like literally airbrushing right um there cannot be a delay between those Strokes and that's that's been another major uh issue with this kind of the resolution is getting bigger and you've got HCR and you've got yep and you have to manage undo and all the database and everything um and it can't there cannot be a lag and now now do it in stereo right depend on pin on both eyes at once now it's double the it can't it can't lag so that's that's one of the reasons why the people you know people enjoy silhouette it just does not slow down it's always better I mean the beauty the beauty of uh the silhouette paint is the number one because it's raster as soon as you let up on your pan it it writes it out instantly so there's it's there's never a build up but I think the Simplicity of it is that when you're painting it's recording everything you're doing every stroke every brush change and once that those series of Strokes are recorded and that means like by time you pin up you can then repurpose those so essentially it's like having a macro being written can constantly so that once those Strokes are there you can then say you know these Strokes I'm going to track those over the rest of the frames because all that work I did on frame one with with good tracking it's going to be good on all these other frames uh and I think just that that that basic recording a macro all the time uh kind of concept and being able to repurpose that on other frames is what makes Silhouettes paint so versatile and and fast on top of it the other the other thing is this goes back to the original design too we've had one of the requests from some customers was they wanted to start the paint but they didn't have the final platelet yet um they they hadn't gone through the degranting or the color correction they were wanted to pre-color correct it so they would do the pain and then get a new plate in and then it would there's a rebuild button it'll go back and redo everything you did from the beginning very beginning uh on the new plate and they've used that a lot you know it saves you so much time so they don't have to wait on the final plate you know because you're not even we're not just talking clown paints here we're talking about painting color yeah and so it's going to rematch yes of it because if you're picking off of the screen if you're picking off the image doing repairs or whatever then it doesn't matter what the new plate is it'll do the same thing you just you did it'll sample the color it'll do everything it nothing's baked in except for the actual gestures yeah that's really cool and my other the other weird one that is always of pain um as everyone knows it's three times I mean there's a whole Reddit page of VFX artists called they're three times and I don't know if you've seen it but it's quite big and I imagine that is something that you guys have to also look at and handle you know yeah by the time it gets to us it's been re-timed usually yeah this is exactly the kind of thing where probably some places would be you know we're re-timing a nuke and then we're going into silhouette and doing everything else let's just let's just skip that step of that retiming right but I don't know do we have anything we can do it we don't we don't have a optical flow like re-time uh or the Continuum product has like an A version based on like you know older technology it kind of needs to be updated but it would be good to have like a really good retimer I mean with digital filter film works we used to use retimer and I'd set I'd set off the analysis like at night on a on an onyx workstation and come back in the morning to have it done and then of course the like as with AI you know the result with you know Optical flow retiming at that time would be all kinds there'd be weird frames you know in the optical flow field when you get like overlapping motion for instance even now even yeah yeah a lot of your paint work shots that are going yeah are coming to paint out optical yeah but that's what we ended up doing we had to end up you had you know one you spent the night analyzing and rendering and then the next day you'd be painting all the fixes and tearing and rips and you know what not that happened from it and um well I guess an example uh when I'm thinking free time is there's this production type uh um problem where a shot is three times it's double speed but the edit isn't quite locked maybe it's quadruple speed and so the question is do I send it to paint and they paint the original frame but then they're painting four times as many frames as they may need do I take the risk have been paying 200 frames but then if they change it now I'm in trouble my question is is there a way in that situation where you could save the work if that return changes so no because we don't have a retiming solution but um if we did then the paint data could be interpolated because it's all Vector internally uh so yes I think we haven't had that request but um when it when it does come time to integrate or to make paint work with re-timed with changing frame rates it would just be a matter of resampling the pain I think from and I think that probably would work but you'd have to rely if you're expanding the paint range it would it would just be you know linearly interpolating out the paint filling in especially doing motion blur on the paint Strokes uh but that should that would I think that would work I think you're the first person to ever I think uh you know it's like we focus on things though you know we know we can do well and time wise we just don't have time to sit down and research retiming um but I don't yeah like I said I don't think we I mean if somebody had had asked for it you probably would have done something by now but again yeah I would have thought someone on any of the Fast and Furious movies or the reason or the Matrix movies I I think you know the a lot of the bigger facilities you know that houses they've got a pipeline right so they do all this this initial stuff all the retiming or whatever they're going to do and then then it comes into solo ad and then it goes on to someplace else but they always have to make it cool they have to make the call do we do we just do the paint on the return plate or we paint on the original plate and there are obviously costs yes cost risk benefits on both answers there isn't a correct one at the moment that's one but for us the the future for paint we'd like to kind of blur more of the line between raster and vector and what and what you can use your imagination is to what that would mean but you know I think flexibility you know after the fact however that however we can make that work we'd like to make it a little bit more flexible um because I think the the more that we can do that uh we don't want somebody to say well the vector system would be better for my particular shot because of XYZ we we don't want somebody to say that um of course we could argue well you know it's going to be slow if you have too many strokes so um but yeah I think that's the future is whatever what reusability you know being able to modify after the fact which again is normally impossible with the raster-based system um you know maybe different things we can do with key framing we did that a feature recently where you could you could sort of preview what your stroke would look like yeah and then lay down a stroke or two and then if you didn't like it you could sort of undo that and change the settings and then just redo it things like that but yeah more Dynamic support would be good um and the the paint model supports it so that that's going back to the whole discussion of what's what's interesting like that's an interesting problem it's kind of funny because it's always I love working on the architecture I love making it extends extensible and putting new stuff into it but it always comes back to Roto and paint for me in terms of what what's interesting yeah and we all we've talked about those are those are kind of boring things there's something that you have to do nobody talks talks about them but but yet to sort of innovate and to make more people happy and see people artists time there are some interesting problems to solve and it is it is always kind of fascinating when we do come up with some some way of improving them um I think anyway I struggled with paint tools even if you were very similar usual or not it's going to be kind of interesting and because they we all know that issues anyone in 2D my other question is what about non um I guess non-realistic pain so if someone was making a movie like from a scanner uh Darkly or back she's Lord of the Rings or something where they wanted to use rotoscoping and paint in an artistic way maybe they wanted to make everything look like an oil painting and they wanted to get some artists to paint things by hand and make them look interesting um do you have support for that or would it be a case of exporting maps to a composting applique I mean I would say I think a little more work would need to be done but currently we have the ability to fill shapes with color um and outline them and you know and outline them um and you can bring you can bring shapes into paint that's for masking purposes right but you could you could do an overall rush I guess through the mat uh but not automatically other unless you use like one of the cartoon filters which is not or painterly probably what you want but yeah I think I think you'd need to uh I think we need to do a little bit of work on maybe gradients within the shapes probably gradient fills yeah yeah take some of the stuff out of the depth note set yeah we have a depth node that that's used for for stereo where there's a bunch of different gradient fill types um you know to simulate depth people are making their own depth maps yeah that that could that could somehow because we've used the Roto node in a couple different places you know the kind of the guts of the Roto node is used in the morph node it's used in the depth node it's using the power matte node you know various nodes that require shapes um you know it could potentially even be a separate node that has more kind of traditional you know incon paint type features for the shapes yeah we haven't really focused on that kind of um special effect paint stuff but we are getting more requests of that nature now that we're a plug-in into after effects for instance we're getting more like uh oh we had somebody do a shot where they hand painted a bunch of like glowing effects on things you know like uh neon light lightning coming off of moving objects and things where they just went and hand painted that on every frame just using like really glowy colors and then they passed that paint just the paint Strokes through another filter that added a nice bright glow effect on top and then comp that back on so there's ways to do stuff like that because you can have the paint node spit out just just the stuff that you painted just The Strokes with an alpha um and then do things with that but in terms of painterly Brushy type stuff no not not yet and the other last kind of I toss that comes to mind in terms of a shot that you might be interesting you might do in silhouette is seamless transitions stitching two different plates a drone to a crane so that you can keep a continuous move those shots are always very hard and challenging um is that something that you for your feature set would be quite strong for or you can do that now with the morph node but in the classic way that even people were doing with elastic reality back in the 90s it was used a lot for the documentary cuts things like that um seamless transitions between stuff uh but it would be a more traditional you know Roto this Roto that morph it across a couple of frames we do have um you two have a project in the pipeline to make that a little easier to do shots like that you can't use layout and cameras here and not at this time no no it's been on the list for a long time we've had several customers wanting to do bring in geometry and do things like that so we're aware of that that need right to ask each of you um if you wanted to in turn I guess in your role if you um so start with Paul if you want to if someone wants to start out and become an engineer and work for one of the software companies what would your advice be to them I I got into computers and things when you nobody you know back in the day when not everybody had a computer right so I had this I was the kid that liked to take things apart and figure out how they worked from a very young age and when I discovered computers that just kind of just changed my world and and so I learned everything I wanted to myself because well by the time I got to college um you know they were still teaching like Pascal uh I was already you know programming Amigas and C so it's just I I was sort of like way ahead of the Curve and I think that voracious appetite to sort of like understand things from a like how the gears work how things work uh was kind of invaluable for me personally and I got to the point where I just wanted to write programs all the time like I want to just make stuff and I think having that sort of that Creative Design urge desire or whatever to create your own stuff um is is probably the number one thing I whatever I'm talking is like let's say a younger programmer I'm like do you program for fun that's always my first question do you do it for you and to me that's more that's interesting I'm thinking oh yeah I I'm I'm always I'm learning some new language I'm running some of these tools I'm doing this because I mean that to me that's interesting rather than just uh you know a nine programmer for nine to five right for instance um so I I don't know if that's that's the more of like a philosophy uh philosophy like an attitude right I don't know if you can teach that it's sort of an innate trait right um that's what I would always look for uh if if aside from all that um whether you have that sort of innate desire to do that or not I I would say if someone wants to get into this industry just I think it really helps helps to unders try to you want to learn what was going under the hood I think that's always really fascinating like um to find out how things work under the hood like learn how the GPU Works how memory works right how how images are represented in memory um and then um well obviously there's the whole languages I assume you're talking about from the programmer point of view right like learning to be a programmer on the bfx industry just I mean there's so much as well um different disciplines you know there's yeah we're doing predominantly 2D type stuff image processing right then you have your 3D all the math involved with that VR uh you know the AI uh even like more uh traditional programming of like you know Asset Management systems and databases just you're just working with data and models and converting formats um I think I'm kind of rambling a bit but I think Learn Python python is like it's the glue language of the whole of our industry and um I felt I discovered it in 1995 and just fell in love with it um I try to do everything that I can in Python uh if I think if you know python you can probably get a job almost anywhere uh in the industry doing at least that um but certainly uh all the major VFX tools are written in C plus there's a lot of python for the glue um and then just you know learn the the various uh tools and plug-in architectures one of the one of the kind of interesting ways that you could get into if you have X industry or and even the even just the the special effects image processing industry is learn ofx you know learn learn the open VFX format and just write a write a filter yeah and then download your favorite package and just see how your filter interoperates with without the software silhouette nuke I mean all the all the big uh VFX packages support of X plugins start with that um yeah I'm not sure if I answered your question but uh it's a stream of Consciousness kind of thing I think you did I think you did I think that's it really no no that was that was a kid um and I guess Marco you're it's a little bit more complicated because I think you've you've done your career path is is more varied in in one sense you've obviously been an artist and an editor you've done compositing you now lead a product so my my question is what is your advice for people starting out um and what is your advice for people moving to different um Fields within the industry I think the the best way uh say for a student is to pick and choose the proper internship potentially at a company that they would like to work because if you can get your foot in the door and you work hard you're not an ass you get along with people uh they'll recognize that and there's a good chance you'll be hired and I think you can't be afraid to start at the bottom I mean when I started I mean I went to film school and you know I had various skills but I took that tape operator job working at night you know not making a whole lot of money but it led to being able to learn their editing system and then eventually editing and then going out on shoots because they had a production arm so don't be afraid to start at the bottom but I think for visual effects I I still think getting in as a roto or paint artist starting there can you know lead to being a compositor if you're talented and not everybody is you can create a great reel and that reel will get you hired any day of the week um provided the facility uh has openings but if you're you're only personally talented but a really hard worker and you can pick up technical skills like Roto um that's a way in the door to being a compositor um you know editing is so decentralized now that's that's kind of a harder I think a harder nut to crack other than again if you have a really good reel um you know you at least have something but it's always about knowing somebody so that that's why the internship is good um you know for me when I went to school there was a group of us in our class that you know kept in contact and when one would get a job on a particular show they'd drag as many of us over that they could and um you know those relationships are indispensable it's like even for me getting I never envisioned being in the software industry it just so happened I showed up at a class and somebody made a comment and I'm like okay you know I I just had the Good Fortune of kind of switching before each industry died you know from one to another um you know when I was in visual effects you know you could make quite a bit of money and I'm sure there's people that still make quite a bit of money now but I think the general artist rate is much lower than it used to be you know I used to work with clients and you know we could you could you could make more money doing that but um again the most important thing getting in the door having a reel if you can create one and uh you know those internships are super important not not that everybody can get those if they're not in school but you know learn the tools of whatever facility that you want to work at and uh you know there's Roto jobs show up at ilm at you know various facilities they may be in a different country so you have to be prepared to move these days um so yeah don't don't be afraid to start at the bottom work hard make a good reel and uh you know it's just like anything it's very competitive and if you want to Lance you know what you can get a personal learning Edition there's a one or a limited uh you if you're a student you can uh you can get a license a complimentary license um otherwise there's a 14-day trial and that's about it but the good thing is is that all the Boris FX products are uh can be monthly subscriptions now with no long-term commitment so you can rent it for a month um you know and there's a lot of really good classes out there there's um you know various companies have classes that you get access to the software during the time that you're taking the class so those are always good to um be a part of but yeah just really learning the tool because the you know whether it's nuke whether it's silhouette uh you want to be able to make sure that you don't have to think about how to use that tool you know it in and out you may not use it well but at least you know how to use it and I think that's the important part absolutely yeah cool and as we sort of come to a close um are there any sort of final thoughts you would like to that I haven't asked you about that you really want to get in before we go as a kind of closing um okay just maybe going off of what Marco said just now I would like to reiterate that I think being able to especially I've assuming you're probably talking to younger generation of VFX sometimes I mean our audience is varied I mean our audience is I mean when when it started Kofi was very much thinking how do you encourage people that want to get into the industry and we are keeping that alive at the same time we're aware that a lot of people who are quite senior are listening and that's why we like to go deeper and get more detailed in and perhaps ask some more difficult questions that might be unclear to some of the English um those that are not quite in the industry yet but um okay but you know yeah I would say for like the you know is going going back to the the points that Marco made uh certainly I think just being willing to put yourself out there and uh one of one of the one of the favorite things that I that I learned from being a programmer and knowing other programmers is is don't don't necessarily if someone asks you if you know something it's okay to say no but just just say I'll find out and then go find out um that that will I think that's probably the biggest piece of advice I can say and I guess I think that probably applies to any age just be willing to go find out and and uh it's okay if you don't know but have that Curiosity as I mentioned earlier to go to go find out and figure it out that'll that'll open many doors I guess I would CL I would close with saying that um I think success in this industry is uh really dependent on just an incredible eye for detail and you can't be afraid or can't be offended about redoing something because doing shots doing visual effects you're gonna redo things over and over and over and even doing software you know we're constantly changing it's you can't get upset about that kind of thing so you you know be willing to change go with the flow and great eye for detail because that's what will matter yeah that's fantastic thank you so much for your time this evening or this day for you guys I guess[Music]