We have another beta available for your testing. This will be released on Tuesday along with the Iris model, so please let us know of any issues you may have.
Adds “Recover Original Detail” option to some enhancement (Iris, Proteus, Artemis) filters
Fixes issue with the export
All audio streams should now be copied to export
Misc updates
Known Issues:
Videos with mismatched metadata and streams will display incorrect duration
Frame number preview length may shorten on app restart
Recover Original Detail
Purpose of this slider is to retain details from original source if the model’s output is overly smoothed
It adds the source texture details to the model’s output to bring back the original details
Lower value might make the output overly smooth and a higher value will also bring in source noise, compression artifacts, etc.
Be careful when adding source noise as this might also show up on output if recover original detail is used
Please look at the tooltip and let us know if the functionality of the slider is understandable from the name and the tooltip
Enhancement model - Iris
Primary aim of the model is to improve human faces
This model is also designed to enhance low- to medium-quality video, particularly, with noise and compression artifacts. The model tends to lose details on high-quality videos
This model is trained to work on all three categories - progressive, interlaced, interlaced progressive - but the main target is progressive videos
Since the model has significant differences from the Proteus model in terms of aim, scope, and performance, we have decided to give it a completely new name
Please test the model and give us your opinion.
Thank you for testing
Please upload problem videos and logs here: Submit Files .
I really liked the old Gaia HQ model since it added a lot of detail. Back in the day, I even used Gaia HQ on very low-resolution (320 × 240) videos, despite stuff looking wonky, specifically because of the extra detail and textures compared to other models.
On my 5 seconds previews, with Iris selected & Recover Original Detail set to its default of 20% it looks truly stunning. I’m now running my usual 50 min test video of 720 up to 1080 from 50fps down to 25fps (Chronos Fast) & Iris Auto and Recover OD default 20% selected and it’s rendering at 5.2fps on my Mac Studio Ultra 64Gb. That’s cool I guess as long as the final output file is better than it was under my previous test of the Artemis model (and all the above apart from the model) which rendered at around 18-20fps. I’ll keep you posted on my opinions of the Iris slower render compared to the Artemis way faster render once it’s finished in a few hours time… so my next post may well be in the morning now, it’s late am UK o’clock y’see!
I really dislike the new Trimming system, it’s broken. Can we go back to the previous system of setting start and end points on the main video timeline, instead of pressing Trim and being taken to a second video timeline to drag start and end points around. I dont know why it was ever changed. We can’t advance or backtrack start and end frames/time using a keyboard and this feature needs to be fixed so we can. This has been an issue since the 3.0 release and mentioned by others.
Been awhile since I’ve posted mainly because the last few months of updates have been disappointing. Every enhancement seemed worse to me than the original because of the plastic facial output until this new Iris model.
Finally I see some normality to the face and I look forward to further testing. I did have to do a ‘relative to auto’ adjustment to properly upgrade according to my own preferences but it was great to have the ability to do so (as with Proteus).
I have been resorting to using “ffmpeg” to split movies up in to individual photos, then batch process them through Photo AI with Face Recovery, then reassemble to a movie using ffmpeg and then use an editor to put the sound back in to it … to get this. (I am just now finishing up on a set of videos that has taken over THREE WEEKS to do this way.)
From initial previews, this looks like it would have saved me three weeks. I will be re-running some videos I already converted manually through it, to compare, tonight.
I’m noticing that the Iris model still has a big issue that has plagued all Proteus models: partial blending of the previous frame during scene changes.
It’s generally gone by the next frame, but it shouldn’t happen to begin with.
I’m also noticing some ghosting, like when a person turns around quickly, and it looks similar to the partial frame blending between scene changes, so I have to assume whatever is causing it, is related.
After testing this new version, I found out I can generate short previews without any issue (albeit without sound) but when I actually press export I am seeing generating preview taking forever and I am unable to check the intermediate results at all.
Is this normal behavior?
I’ve generally notice this happen when the output video has a non square pixel format, but I don’t know if it occurs only then, or if there are other underlying reasons.
Trimming still doesn’t trim the frames you ask it to. Frames 33300 to 33400 I just tried generated 33299 to 33398.
The exported frames are one digit off again. The first image below is the scene boundary of the exported trim, the second image is the scene boundary of the input, indicating the export is legitimately starting at frame 33300 but naming it and the rest of the sequence starting from number 33299, putting the sequence one frame behind on export.
Even if the name was corrected, it finishes then on frame 33399 not 33400 as per trim so it dropped a frame.
Your still trying to play the exported files as the preview, likely still why issues with preview. If you export over time, Preview breaks again still.
Initial results mixed. Seem very similar to previous version though I am noticing some wobble I have not seen before in Proteus/Iris. This is a slider of original, last Iris V1 and current Iris V1. Relative to Auto, 10,40,50,24,10,40, no noise, 20 on new slider, grain 2/2.
Ill keep playing with this and the new slider and see how it goes.
Recover Original Detail doesn’t seem to work correctly with Proteus V3.
In Iris, Artemis and Proteus V2, I can see an obvious difference between 0 and 100% (and also much lower values) but for the same content I can’t spot any difference when using Proteus v3.
Did some more tests. IF the topaz devs can increase temporal stability, so that objects retain the enhanced details despite something or someone walking in front of it, and most importantly FIX THE BLENDING/GHOSTING ISSUE, then I think Iris v2 or v3 will be very good.
FIX THE BLENDING/GHOSTING ISSUE 
, then I think Iris v2 or v3 will be very good.