Fixes issue with estimate not working for certain image sequences
Updates NVENC (NVIDIA encoder) - this requires at least driver 531
Beta build for AI Engine 3 (No user facing change)
Updates for Iris model from previous beta
Known Issues:
Videos with mismatched metadata and streams will display incorrect duration
Frame number preview length may shorten on app restart
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
Please upload problem videos and logs here: Submit Files .
The image quality is quite good and dependend on the source material maybe the best so far. Currently testing a 1280 * 544 video upscaled to 1920 * 816 with Iris V1.
Getting what is probably corrupted output files from this build. Using VLC, video stays still and audio continues to play. Happens with every source clip I’ve tried so far (using Apollo and Iris, 2x upscale).
Logs and sample clip uploaded. Also uploaded bad output file.
Edit: tried more than one output format and made no difference.
Edit #2: the output does play properly on windows media player but does not on VLC. Something has to be off though, VLC usually plays everything
DPX sequences consist of many frames, often at high resolution. To export and edit a DPX sequence, you need a computer with a high processing speed.
how about this
You must pay a license fee to edit in DPX on post-production and film editing programs.
The file size was the same for the one processed yesterday with v3.2.9b and the one processed with v3.2.10b.
Is there anything different in the “Beta build for AI Engine 3”?
Yes, this does appear to be model specific and we are working to determine if it is fixable with the current Proteus model or if a new model needs to be created.
Yes, this happens most often with high res files and can be quite annoying. The preview experience is a major project and we have run into a few roadblocks along the way.
Face improvement is a very good thing, but, if it’s going to lose details in other parts of the image, then it’s really not worth using IMHO. If there could be a model that improves faces AND enhances detail… I think that’s what everyone really wants, isn’t it?
I’ve been thinking about how to best improve the iris model.
Iris does 2D faces amazingly well as if follows the image (Faces) on the screen and produces a result from exactly what it sees, so if an eye looks misaligned due to compression artefacts it will match the eye’s position no matter how bad it messes up the face.
I’m guessing that it has to do with topaz using a similar model from their image rescaler, which is an excellent way to apply it here for video, except video features 3D faces in 2D, what we need is iris trained with simple geometry as well as 2D facial reproduction, say an egg-shaped oval for the head, that way it can predict the face’s features position such as eyes, nose, mouth on a 3D model.
Visualise a projector projecting a face on a 3D model of an egg, and you’ll see what I mean.
Currently, uninstalling the previous and installing this beta now, I’ll give it a go.
EDIT: It takes a bit to uninstall as I’m using windows uninstall troubleshooting prog, as I cannot uninstall this program due to that accursed permanent deleting shortcuts bug!
