I’ve done some tests on the Chronos model using the three video quality Settings; Low, Medium and High (all other settings the samel) and the resulting file sizes are identical. While it’s not impossible that different processing would produce the exact same file size, it’s highly improbable for three, suggesting two possibilities;
The Chronos model does not use this setting and all videos are processed the same
There is a problem with the Chronos model and this setting and the outputs should be different.
I can see why the former would be the case as generating intermediate frames should have little to do with the quality of the media, but want to eliminate the latter as a possibility.
Thanks in advance.
viktorz3008
(Z...z...z 😴 I'm sleeping. And I love Sumire 😍.)
2
Do we have video quality settings? Where did you get that thing?
Top Right, It’s the very first thing used to guide the model Suggestion;
Video Quality
Low | Medium | High
Video Type
Progressive | Interlaced | CG
Video Artifact Type
Motion Jitter or Choppy Framerate
viktorz3008
(Z...z...z 😴 I'm sleeping. And I love Sumire 😍.)
4
It’s not the quality of the output video. It’s for you to judge the quality of the input video. If your video has low quality, choose Low. If high quality choose High. Different models will give different results.
You specify the source parameters and it suggest a model. With all the others, the Low, Medium and High produce different ‘sub-models’ (is that the best term?), so for Artemis you would get Artemis_LQ, Artemis_MQ and Artemis_HQ.
Since I don’t know what goes on under the hood, it’s not clear what these ‘sub-models’ are; is it a different algorithm or is it the same algorithm with different training data.
Chronos just has the one result.
So your suggestion;
…makes no difference in this case, the video is processed the same.
Is this correct?
viktorz3008
(Z...z...z 😴 I'm sleeping. And I love Sumire 😍.)
6
Artemis HQ, MQ, LQ are exactly what you think they are.
But for other models, the output would be the same everything because they’re just a model without any modifications.
If you want to adjust the sliders, use Proteus model.
As far as I have been able to tell, the options that you are asking about are only there to assist you in choosing an AI. They make no difference in how the AI works. If you look at the all models tab, you’ll see everything available to you. A/B testing and using the compare view are helpful for quickly finding the best one for your video.
You are correct. Chronos only interpolates frames. You can either upscale before or after, although both have their tradeoffs. Chronos doesn’t seem to handle duplicate frames well either, so you should use ffmpeg or a tool like FlowFrames to remove the dupes first.