you’re late. Topaz is actually reducing all their pictures and Ai suite in Two only software. Denoise + gigapixel + sharpen = Photo Ai and Video Ai separated, but Photo Ai is not ready yet to totally replace the whole suite, there are still stuff missing but it’s going in the right decision i think.
i hope they’ll stay away from subscription, it’s the main issue with Adobe.
1 Like
That’s a plus and thanks for the information.
1 Like
You’re welcome 
I am really hoping that add a model that can clean up motion blur from videos that were recorded at too low of a frame rate for the motion they recorded.
1 Like
Well, halloween is over, and monster faces are out. I cross my fingers that TOPAZ video devs get capable to deliver AI enhancement that could compete better with what is available with single image AI enhancement. Off course there is always a compromise between acceptable speed and reasonable quality. For the time being I’ll stop beta-testing of V3 because actually it’s just a waste of computer resource.
Hi - are you running into consistent crashes with any video? If you can share with me an original file you run into the crash with through Dropbox, please. If you can also turn “Enable Detailed Logging” on in Preferences, reproduce the crash and send me the logs from Help > Logging.
Thank you,
1 Like
Changes to address these can take some time, but we’re already looking into the performance and model quality, and many other things. We appreciate your feedback.
8 Likes
Hi Yazi,
Thank you for your answer.
Yes, it happens with any video (and it never happened before v 3.0.2).
I downgraded to v 3.0.0 so that I can continue to work, so it won’t be possible to send you a detailed logging right now, but I’ll do that ASAP.
1 Like
Thanks for your answer. You integrated old models too ? As Artemis 7,8,9 and others ? Thanks
Wow! Where can I find the tool to combines CodeFormer and Real-ESRGAN for my videos (like you said for upscaling people’s faces + backgrounds)?
I searched CodeFormer in Github, there is “Video Enhancement:” section, is that what you are referring to?
1 Like
Yes, you can download the code from Github and AI models from Google Drive, then follow the instruction to install.
After that, you can use the command such as below to upscale video file (mp4, mov, avi).
python inference_codeformer.py --bg_upsampler realesrgan --face_upsample -w 1.0 --input_path video/video.mp4 --save_video_fps 24
It will automatically, extract each frame from the video and search for faces. Face will be upscale with CodeFormer model, while the surrounding will be upscale with RealESRGAN. After all frames being processed, it will join them back into a video file.
1 Like
Thank you! Please let me know here or through a private message once you’ve got the chance to send the info. I appreciate it.
Sorry. I tend to be tactless if I don’t check myself. EDIT: I see Yazan has already said what I was going to reply.
@jam7000, thanks for your comment. Can you please share some sample videos with the motion blur that you are talking about? Submit Files in DropBox. Thanks.
@bug008, seems like the video has non-square pixels. Try changing the “Original Pixel Type” to “Square Pixel”. If still doesn’t work, please send us the video: Submit Files. Thanks.
I don’t understand,why the version 3.0.0.8 is very fast for encoding and the official release is very slow. It’s strange for me. 
Explain me please ?
@nipun.nath I can only now produce a min of 60 frames as TIFF in trim.
Hi,
We’ve updated the minimum driver version in order to add support for AV1 encoding on supported cards as well as bring some performance improvements we’d noticed internally.
While older drivers may’ve worked in some cases, we can’t guarantee that bugfixes or other changes work with them as we don’t test against drivers below the supported version.
1 Like
@gregory.maddra In Trimming I can only export a minimum of 60 frames.Also Can cropping be also adjusted by the a on screen grid like Photoshop 
@yazi.saradest
Img0026VEAI303MediaInfo.txt (11.3 KB)
I just noticed the BETA will give a warning and refuse to upscale beyond a certain width/height size. I can take a vertical video and rotate it, and then it will handle it. Some MP4 limit?
But, on the old 2.x.x version, it handles the video and exports it, just with a warning that it might not work.
Since my video res X-by-Y did not work but Y-by-X did, is this something the program could just handle automatically?
