Hi y’all! I see lots of great feature requests on our forums. Starting with v 2.3.0, you can now submit feature requests directly through the Help menu in the app.
This will be the best way to get your feature request counted. While we can’t guarantee every request will be included, we use this information to help gauge user interest and inform continuing updates.
Please keep posting your ideas here as well because they lead to some great discussions and you might find that feature you were looking for already exists or a clever workaround as well 
Right now this is only in Video Enhance AI and Gigapixel AI however we may add it to other applications as well.
I’ve closed other feature request threads for now but feel free to comment below if you have any questions!
Emily,
The new Proteus model is awesome. I’ve been able to get better results far faster and easier than ever before. - However, that raises another question. I frequently need to pre-edit the files I run through VEAI. What is the optimum file format to feed Proteus? Would using Dynamic GOP help? There must be some input file factors that could help or hurt VEAI do its job, especially Proteus.
Any comments or relevant links/info would be greatly appreciated.
Thanks,
PDWhite
A Model or multiple models that improve the colors
for example:
SDR to HDR conversion
Black and White to color
Automatic color correction (based on reference image or video)
Linux support
It does already work with Wine, but Cpu only and is very slow. A native port with full gpu support would be great.
I’m a CG Artist currently testing 3Delight which doesn’t come with a built-in denoiser.
I know there’s Intel Open Image and Nvidia Optix Denoiser out there, but the result of these 2 often softening the final result too much.
I’m wondering if VEAI has any plans to support 16bit and 32bit multi-channel .exr image sequence? so that we can denoise all of the contribution AOVs in one go? and also would be great to have a batch render option to denoise multiple render layers simultaneously.
Hi! In general, file format doesn’t make a difference in quality however, if you use a heavily compressed format, it may introduce artifacts that can be harder to remove but VEAI can compensate for those as well. The best practice is just to give VEAI the best quality source you can as early in your process as is reasonably possible.
There may be instances where it makes more sense to use VEAI later as well and these guidelines may change in the future so experimentation is encouraged
Hi! Not in the update planned for early next month, but definitely a request we’ve received before and would want to seriously consider.
Quick tip, if you do use VEAI, Gaia CG is a good option but, depending on your source you might actually find that Artemis HQ is great on CG as well!
I have experimented with making two passes through VEAI. The first is to clean up and the second pass is to render up to the desired scale. - This works, but only works really well if you set the (mp4) output CRF to a very low value, like 1 or 2. - This results in a humongous intermediate file size, but the model doing the 2nd pass will have much easier to digest input and makes for better (final) output file compressed at a more reasonable level - like 20.
This leads me to making a suggestion for future functionality. That is to be able to chain the process of a given video clip sequentially through two models. Given a rather uncompressed intermediate ‘temp’ file, the two models won’t ‘artifact’ each other and makes it possible to queue up the sequence to run overnight.
A way to save progress, recover partially completed files? Rendering files with Gaia can take 8-12 hours or more, if I get a crash 70% of the way through all of that work is lost and not recoverable.
Surely there exists a way to allow us to recover these partially processed files and resume progress and then join the two files?
Regarding the Proteus-6 parameter model, consider adding a button that will run the auto feature on multiple frames (e.g. 20 equally spaced frames in a clip) and then average the values. This will help optimize the settings for a single clip rather than a single frame. Being able to set the number of frames analyzed would be of value too.
As it is, I manually average the values in excel.
I recommend you try using ProRes 4444 XQ for the first clean-up pass. The file size is comparable to MP4 with CRF 2, but I noticed that the second upscaling passes are way faster… at least with using the Chronos model to increase the framerate.
[Edit] Nevermind, maybe it was a fluke with my computer yesterday but I can’t duplicate the results… at the moment both are showing to be equally as fast.
Hi there,
I would like to put forward a couple of feature requests, that would make Video Enhance AI actually usefull in animation and VFX studios.
- Proper color management, meaning OpenColorIO with proper input and output color spaces- To ensure all the math is done in linear space.
- OpenEXR file format for input and output.
- Meaningful output naming of image sequences. At the moment I can specify a folder where to put the images and then it renders them named 000000.tif, 0000001.tif and so on. I cannot even specify a prefix or even the amount of digits that the sequence numbering should have.
- Did I mention proper linear calculations before?
- RLM Licensing. Online login licensing might be cool, simple and comfy for the average user, but for closed studio pipelines, where the workstations have no connection to the internet, and for good reasons (leaked footage anyone?), it is a complete no go. There needs to be a possibility where a studio can buy x number of floating licenses and have a license server manage the distribution of the license in use, while the workstations are offline. It’s not a new concept after all.
- Scriptability and headless comandline usage. A lot of our daily tasks at the studio, are automated and scripted, using shell tools. Would be nice if VEAI would do that as well.
- And last but not least. It would be great to have this as a Nuke plugin. Standalone is nice, but being able to upres directly in compositing, would be a huge deal for animation and vfx pipelines, as the available tools just aren’t nearly as good as yours. I know this is a tall order, but I suspect a lot of site licenses if you go this route.
I am baffled at the potential this tool has and the quality it can produce, yet I cannot use it for professional work, because it messes with the colors, due to not being color managed correctly. Resulting in a beautifully upscaled version with unfortunately the wrong colors. Not by much, but that is what you get, when your math isn’t performed in the right color space. It is frustrating to say the least, because I have looked far and wide and have yet to see any professional upscaler for moving images, that creates even half as good a result as VEAI, but I cannot use it on big productions, because it is too limited in the way it operates. Time will tess if this tool will stay in the hobbyist realm, or some day find it’s way into the professional world.
Loving the app so far! However, if I use it to filter certain videos that are of lesser quality, either from film grain, compression, or otherwise, it may sometimes denoise certain sections more than others. Is there a way in the future that we can mark out sections of the video (in frames) where we want the algorithm to pass through without applying any filters?
I find your comments odd. For the most part, I’ve seen very little color alteration using VEAI. However, in a few cases my source material was oversaturated, which degraded VEAI’s results. - In those cases, I preprocess the video by lowering the saturation and bring it back up after VEAI has done its thing.
Another issue *may be * the color standards being used by your editing tools. Most video uses REC. 701. But a lot of software is oblivious to that. You may be just looking at it through sRGB or what a lot of monitors call ‘normal.’ - My monitors utilize an app that changes the monitor’s standard based on what application is currently active. - It makes a big difference.
Look at the display for the apps you’re using and check to see if your monitor settings match.
Yes! Functionality along those lines would be a new level to the current one which is solely focused on one processing one file/one configuration and transform VEAI into a comprehensive video restoration/enhancement tool.
However, it is already possible to load clips and process only a part of them. I haven’t tried this yet, but I think the same original clip can be reimported and a different section of it processed in the same batch.
Regardless of whether this added editing functionality gets added or not, VEAI already performs miracles with otherwise impossible image problems.
Because I am dealing with purely cgi animations imported as image sequences, I know pretty well what my color space is and I have tried converting my linear renderings to all sorts of color spaces, including rec709 and sRGB. None worked, saying all of them got color and saturation shifted inside VEAI. And no, trying to fix the shift in post is not an option since I woud only be able to get in the ballpark but never exactly where I want the colors to be. I’m sorry, but in cg animation and vfx, this suggested workaround is a total no go.
We NEED audio passthrough. No conversion so we can get 5.1 audio, DTS, DTS Master etc. I hate having to separate audio then do the VEAI then merge the audio. So many steps when you can just do audio passthrough
And we need MKV output also.
Can we please get an improved Artemis Aliasing/Moire preset for more complex moire and aliasing issues? Or even a chromatic abberation removal too. I could send more video samples of these. Just please let me know where to submit.
