Chaos Corona 13 for 3ds Max - Daily Builds Discussion

[michaltimko]

The AI enhancing icon is greyed out in the VFB, even after adding the render element. Anything else we need to do in order to use it?

Same issue, tried different scenes. It`s not greyed only if i use IR.

[Aram Avetisyan]

The AI enhancing icon is greyed out in the VFB, even after adding the render element. Anything else we need to do in order to use it?

Same issue, tried different scenes. It`s not greyed only if i use IR.

Hi,

With current implementation (and limitations) both sides of the render should be within the range of 512 - 2048 for AI enhancer to work. No limitation on aspect ratio though.

[maru]

@Bjoershol @michaltimko - was it caused by too high resolution? Is it working now?

[Bjoershol]

Okay, it's working as long as I keep the resolution within 512 - 2048!

[Juraj]

Even my smallest preview renders I send to my clients are 2560px, but usually I don't go lower than 3000px even for that.
2048px is just brutal limitation for 2025 I am afraid.

[TomG]

Remember that the AI Enhancer is also an upscaler - it gives up to 4K in results, accepting up to 2K as input. (And with this being under review - you can always add or vote on the AI Ideas Portal for preferred limits :)  https://chaosai.ideas.aha.io/ )

[dfcorona]

You have "notes" under "CMasking_AiEnhancer". Why not add the resolution limit to stop the confusion between users. Just a though.

[Aram Avetisyan]

You have "notes" under "CMasking_AiEnhancer". Why not add the resolution limit to stop the confusion between users. Just a though.

Good point, thanks.

[michaltimko]

@Bjoershol @michaltimko - was it caused by too high resolution? Is it working now?

Yes, it works now if i keep resolution down.

Even my smallest preview renders I send to my clients are 2560px, but usually I don't go lower than 3000px even for that.
2048px is just brutal limitation for 2025 I am afraid.

Same. Was thinking that i can skip stable diffusion but 2k is quite limiting for many. My finals are between 5 and 10k, mostly in bottom range.

[TomG]

Just a reminder that 2K is the max input resolution, which results in a 4K final image that you would deliver to your client. This will be reviewed as time goes on, but I did do a check on Behance, the Gallery here, etc. and most images were 4K or under so it seemed like a reasonable starting point :)

[alexyork]

Just a reminder that 2K is the max input resolution, which results in a 4K final image that you would deliver to your client. This will be reviewed as time goes on, but I did do a check on Behance, the Gallery here, etc. and most images were 4K or under so it seemed like a reasonable starting point :)

I think it would be best to get the opinions of industry people directly rather than interpreting what you see on social media (which is always lower res than the "originals", for obvious reasons of security etc.). The industry has been at 5k minimum and often higher than this for a very, very long time, for final still images. We've been at 6k for about 5 years, probably more, and many of our clients require 10k, 12k or more sometimes.

There's no way we can reduce the native res of images to as low as 4k or 2k these days just to take advantage of AI enhancements. We'd just be forced to use any of the multitude of alternatives out there (which we're doing already). What if your current implementation could work at the user's high res, internally down-scale, do it's AI "thing", then push it back up to the native high res again? I.e. reduce to 2k then back up to 5k/6k. ? It's still far from ideal but at least it would enable users to work with this.

Confident you guys can figure this one out :)

[lupaz]

Or just focus on the ideas with the most votes?
I know that the ai effort is Chaos-wide, but there are many other things on the list and these features are very much replaceable with other existing and better products out there.
I don't mean to tell you how to do your product. This is just my two cents.

[TomG]

BTW we didn't set the resolution limitations, those were already there in the AI Enhancer :) And we (V-Ray included) raised the fact that for our users, higher resolutions as input and output would be required. So it is on the cards to have this improved. Tests were done on using a 4K input, downscaling to 2K, running the enhancement, then reupscaling to 4K, and in fact the enhancement results are better in this instance than starting with a 2K input. So we will see some sort of change to this over time, probably incrementally.