Huge fan of Corridor. For those that don't have twenty minutes, one of the big takeaways is that a pretty high level VFX Supervisor critiques the keys at the end and basically has only nits that you need to zoom in on or have a trained eye to catch. Very legitimate feedback for a top tier professional context, but this can get you really far otherwise.
It doesn't output the final result that you'd expect from a human professional, but it does seem to be well into the professional level of tools. And for cheaper productions you could use it directly.
I mean, having done quite a bit of rotoscoping work in high school, if you can start from the output of this plugin and refine a little, you’re absolutely still saving hours. It was still most of the way there.
A beta price of $50 or full price of $100 is really a steal. If you do ANY sort of video production work, this is an amazing tool. The more casual of a green screen person you are, the more impressive this is.
This is like Magnet Mask for Final Cut, but probably better. I haven't compared/contrasted the two yet, but I have faith that Corridor has knocked this one out of the park from the demos I have seen.
(Yes, you can download it for free. You're paying for the pre-packaged convenience. I 100% paid them to support the work they are doing).
Obviously this prioritizes quality over speed, but how far are we from real time? Hooking something of this quality up to a live webcam would be a dream
I'd love to use this but sadly my local GPU is just crap. I know everyone hates hosted offerings but I'd be a happy customer; I mostly use topaz's hosted stuff too for this reason
Just to be clear, this isn't generative AI in the same way something like Nano Banana or Seedance is. You don't prompt anything into existence. However, this is using all of the latest SOTA techniques that the slop generative models also depend on.
I think it is an amazing example of how labs should take all of their fundamental research and turn into products entirely designed, limited and scoped only to empower artists instead of obsessed with trying to bypass them.
> labs should take all of their fundamental research and turn into products entirely designed, limited and scoped only to empower artists instead of obsessed with trying to bypass them.
The software companies aren’t going to start doing this because the media companies aren’t interested in doing that. Tools like Nuke that already do this are expensive tools for people that the tech industry is trying to make obsolete. The media companies want to sell anything that will sell, and spend as little as possible in the process.
I’m pretty sure this will have to come from skilled and experienced artists doing amazing things with technology despite the tech industry sabotaging their careers, which in turn develops demand for actual art rather than bulk-created ‘content’. Sadly, I think we’re years away from that. I graduated in this field in 2024 and my career was dead before it started. Most people in this business with the hard-won skills to do this work that spent years developing their taste, storytelling sophistication, and craft are in the process of being systematically fucked over by the tech industry. Sad.
I suppose for version 3.0 they will abandon the whole green screen thing completely and the model will just recognize by itself (possibly with a user supplied hint mask) what the background is that needs to be removed and what the foreground is that should be kept, without any need for green screen or blue screen footage.
I think there is still a little bit of room for green screen. At least maybe until Seedance 3.0 in a few months.
But it is interesting how video generation turned the corner from several months ago to now. I think the next AI project they do probably has to be with a generative tool even though they may be very hesitant to do that because they don't want to emphasize AI models that threaten their own jobs.
But I remember when they did the Project Greenlight thing which resulted in a massive list of desired projects from subscribers that they could not possibly make a dent in.
I think in the next several months they may be able to revisit that or at least some of the projects that had to be shelved due to lack of time. To make that possible they would leverage the latest generative AI models and workflows, possibly a few that haven't been released yet. But we are rapidly approaching the point where the models plus tooling is good enough to replace 95% of the technical work, and it's down to creativity. I think it's only a couple of years before that progress bar really maxed out.
Indie filmmaker who had an original animated feature film in Sundance featuring lots of green screen in it many years ago- green screen is dead- game engines + pure cg/2d anim are the future of indie film-
Directors hate green screen, actors hate green screen- obsolete visual fx artists larping as storytellers/filmmakers love green screen.
I hope they open source it before someone else decompiles it and does it for them. Just charge for binaries that most will pay for anyway. Then things like Linux support could be handled by the community.
Attribution-NonCommercial-ShareAlike 4.0 International Public License
(CC BY-NC-SA 4.0), the full text of which is included below, subject
to the following additional terms and conditions. These additional
terms supplement the CC BY-NC-SA 4.0 licence and take precedence
where they conflict.
"This tool" means the CorridorKey software in this repository and
the CorridorKey model weights distributed for use with it.
By exercising any rights to the Licensed Material, You accept and
agree to be bound by both these Additional Terms and the
CC BY-NC-SA 4.0 Public License.
1. PERMITTED USE
a. You may use this tool for any lawful purpose, including for
processing images as part of a commercial project, provided
that such use complies with the restrictions set out below
and the terms of the CC BY-NC-SA 4.0 licence.
b. The images and video You produce with this tool are Yours.
The Licensor claims no rights in them, and You may use them
commercially.
2. RESTRICTIONS
In addition to the restrictions set out in the CC BY-NC-SA 4.0
licence, the following restrictions apply:
a. NO REPACKAGING OR RESALE
You may not repackage, redistribute, sublicense, or sell
this tool, in whole or in part, as a standalone product or
as part of a competing product, and You may not charge for
access to this tool's code or model weights.
b. NO PAID API OR INFERENCE SERVICES
You may not use this tool to provide inference as a paid
API service, whether directly or indirectly. This includes,
but is not limited to, offering access to this tool behind
a paywall, subscription model, or usage-based billing
system.
c. DERIVATIVES MUST BE FREELY ACCESSIBLE
You may build on this tool, but any project, fork,
derivative work, or modified version that You release and
that incorporates this tool must be made available to the
public free of charge. Voluntary donations and
pay-what-you-want pricing are permitted, provided that
paying nothing is always an available option.
d. COMMERCIAL SOFTWARE INTEGRATION
If you operate a commercial software package or inference
service and wish to incorporate this tool into your
software, you must obtain a separate written agreement
from the Licensor. Please contact:
[email protected]
3. LARGE ORGANISATIONS
a. If You are, or are using this tool on behalf of, a Legal
Entity whose total gross revenue exceeded US $20,000,000
(twenty million US dollars) in its most recently completed
fiscal year, the permissions in Section 1 do not apply
unless that Legal Entity has purchased a seat of the
CorridorKey plugin from the Licensor for each individual
who uses this tool.
b. "Legal Entity" means the entity using this tool together
with all other entities that control, are controlled by,
or are under common control with it. Revenue is measured
across the whole Legal Entity.
c. Each seat of the CorridorKey plugin includes a licence
for one individual to download and use this tool and its
model weights. Seats are available at:
https://corridorkey.com
Niko goes into some more detail about the licensing choices they've made and why towards the end of the video as well.
[Edit: note that the above video is "content" so if you're not used to the style it might be offputting but I'm into it]
It's been so cool to watch this develop
Source: https://github.com/nikopueringer/CorridorKey2.0
This is like Magnet Mask for Final Cut, but probably better. I haven't compared/contrasted the two yet, but I have faith that Corridor has knocked this one out of the park from the demos I have seen.
(Yes, you can download it for free. You're paying for the pre-packaged convenience. I 100% paid them to support the work they are doing).
The commercial package primarily consists of commercially supported plugins.
I think it is an amazing example of how labs should take all of their fundamental research and turn into products entirely designed, limited and scoped only to empower artists instead of obsessed with trying to bypass them.
The software companies aren’t going to start doing this because the media companies aren’t interested in doing that. Tools like Nuke that already do this are expensive tools for people that the tech industry is trying to make obsolete. The media companies want to sell anything that will sell, and spend as little as possible in the process.
I’m pretty sure this will have to come from skilled and experienced artists doing amazing things with technology despite the tech industry sabotaging their careers, which in turn develops demand for actual art rather than bulk-created ‘content’. Sadly, I think we’re years away from that. I graduated in this field in 2024 and my career was dead before it started. Most people in this business with the hard-won skills to do this work that spent years developing their taste, storytelling sophistication, and craft are in the process of being systematically fucked over by the tech industry. Sad.
Just go full cg or full ai gen.
But it is interesting how video generation turned the corner from several months ago to now. I think the next AI project they do probably has to be with a generative tool even though they may be very hesitant to do that because they don't want to emphasize AI models that threaten their own jobs.
But I remember when they did the Project Greenlight thing which resulted in a massive list of desired projects from subscribers that they could not possibly make a dent in.
I think in the next several months they may be able to revisit that or at least some of the projects that had to be shelved due to lack of time. To make that possible they would leverage the latest generative AI models and workflows, possibly a few that haven't been released yet. But we are rapidly approaching the point where the models plus tooling is good enough to replace 95% of the technical work, and it's down to creativity. I think it's only a couple of years before that progress bar really maxed out.
Directors hate green screen, actors hate green screen- obsolete visual fx artists larping as storytellers/filmmakers love green screen.
They’re charging for the bundled plugins for various editing suites
"This tool" means the CorridorKey software in this repository and the CorridorKey model weights distributed for use with it.
By exercising any rights to the Licensed Material, You accept and agree to be bound by both these Additional Terms and the CC BY-NC-SA 4.0 Public License.
1. PERMITTED USE
2. RESTRICTIONS 3. LARGE ORGANISATIONS Etc. Pp.The OSI was created by the hyperscalers to pilfer free work from the commons.
Redis and elastic got owned by GCP and AWS because they went for OSI purity. The OSI is the hyperscalers.
Good fair source is more like shareware. The community can do whatever they want. Companies trying to profit must give back or pay to play.
Developers need to be compensated. It shouldn't be a labor of love that gets commercialized by hyperscalers whose own products are not open source.
The Corridor Crew model here is fantastic. As are fair source licenses. That's the way we should have done things all along.