logsForSupport.zip (1.3 MB)
DxDiag4_30.txt (104.5 KB)
Passed the system check, but not able to render anything yet. Any guidance is appreciated.
Single-byte character filename.
Did a new version of Starlight Mini just get pushed out? I was running it earlier and after a reboot and relaunch of 7.0.0.1b, Starlight asked me to run the system check again and started downloading once more. When I downloaded it the first time the model was 6GB like the photo above, but the one that is downloading now is 4GB>
additional note to the above failure to work, the system check screen and download bar stays on the screen after the download is complete. It will go away if i click on anything else.
8-bit max names, no special characters, I wouldn’t use underlines either. The error you get, is the same now?
yes, error is the same.
But i think its the input resolution, its a full hd video file.
I want to use starlight mini but I’m on a mac so I can’t access it, it’s great.
I am using a 12GB 3060 and the plug-in is working. Source is 720x406 (widescreen SD) saved out of Resolve as Apple ProRes. Upscaling 20 seconds by 2x took about 2 hours. I thought it wasn’t working because it took ages to load the model in the queue, but then it started encoding.
I chose to make the results as a TIF image sequence because I can see an image immediately and if it goes wrong at some point everything will be saved up to that point. I will combine after in Resolve. It is comparable times to rendering a complex scene from Blender.
I’m getting an export error on an RTX 3090 Ti system with 64 GB RAM. After analyzing the log with a chatbot(since I couldn’t understand anything myself), here’s the diagnosis:
The export error is caused by a Python 3.12 incompatibility with the CFFI library used by
pycryptodome, which leads to a failure inrunner.exe. This prevents the generation of the intermediate video file, causing the subsequent FFmpeg process to fail due to a missing input file. The issue is further complicated by actypesAttributeError, suggesting a broader problem with the Python environment.
Starlight is super slow, but maybe there is more here. Once I had a case TVAI stucked and given me absurd high rendering time. I was SD 16:9 none codec friendly something like 853x480. Then I added with virtualdub black bars arround it fits 864x486 (<-is ecactly 16:9) and TVAI had no issue with that. I’m not saying that’s the problem in your case, it could be something else, but i’m having deja vu by reading this. Source that create floating points is never good.
If you made this try, go with 720x406 to 736x414
2 Likes
Make sure your CUDA fallback policy allows the model to be spread onto RAM when it is too large to be contained in VRAM. I think the load will fail silently otherwise and get stuck in “Loading Model”.
1 Like
New thread here:
Hi, would u like to share some input and output samples with these low quality eyes and waxy skin? long videos is not necessary. My email is xiaotian.li@topazlabs.com. I appreciate that!
