I have briefly tested Whisper locally, basically following the instructions: https://github.com/openai/whisper, I’m very satisfied with results and speed, but not sure where I can use it yet.
Flux is a very hot topic now, it’s better in some cases than SD3, so it took more time to test Flux in ComfyUI, generally I had to follow suggestions here: https://github.com/comfyanonymous/ComfyUI/issues/4165#issuecomment-2264948167 to downgrade torch in ComfyUI installation (fortunately it was easy with venv, just be sure you are in env). Still the results are not exciting because of generation time on my Mac – above 500s per image, if I could only switch to some more robust machine. Here are some examples:
