tazztone

Results 34 comments of tazztone

shouldn't it be possible to **losslessly** save a clip (60fps, 1sec) to a clip (30fps, 2sec)? all the frames would remain the same, just be played over a different amount...

tried some topaz photo AI face restore and sharpening on that and might as well share it ![TPAI_Clipboard_26f1eadbfce1472c31cfd6e33dcc6fc8-topaz-faceai-sharpen](https://github.com/user-attachments/assets/128d307d-1dda-4c0b-af86-57308f93e26b)

> What OCR language(s) do you have installed, and which one were you using to do this extraction? > > /needinfo those 3: Name : Language.OCR~~~de-DE~0.0.1.0 State : Installed Name...

> maybe it has something to do with the display settings not properly translating the coordinates. that's not it. let me demonstrate by using text extractor on this: ![2023-08-30 19...

i do have "ditto" installed. (EDIT: i get the same Result when Ditto isn't running in BG) and i can make a GIF as well if u like me to...

sure, here u go: first one is with shortcut (powertoys). 9/9 times it only captured "crop : ". with textgrab.exe it worked 4/4 times. ![Recording 2023-08-31 203024 x (2)](https://github.com/microsoft/PowerToys/assets/62671577/64c50406-dc38-4647-a351-31c0b6a0397d) (tho...

i observed that when lora is added to nf4 flux prompt the vram usage increases from 12 to 24gb during generation. and takes forever to finish. but after that, subsequent...

also found this unpredictable behavior of loading/unloading of models on my 3090. maybe some checkbox to "keep model in vram" would help?

since the last couple of changes it seems to have gotten better. i can now load flux q8 and t5 q8 and keep it in vram between generations ![image-19](https://github.com/user-attachments/assets/ce8234cd-d962-4aba-ba60-0465da1563b9)

set this to skip lora patching ![image](https://github.com/user-attachments/assets/0a77f65f-db58-40ad-82e8-e0de2d458d06)