When a character is loaded and a PFP is generated in chat.py > generate_pfp_cache a thumbnail is generated as well.
This thumbnail image is then returned back to state['character_picture'] for us in Gradio.
When saving the character, state['character_picture'] is saved as the image parallel to the character .yaml, overwriting the original, and potentialy larger, image.
If this was the intended behavior, I would have assumed the down scaling would be performed when an image is initially imported.
Describe the bug
When a character is loaded and a PFP is generated in chat.py > generate_pfp_cache a thumbnail is generated as well. This thumbnail image is then returned back to state['character_picture'] for us in Gradio. When saving the character, state['character_picture'] is saved as the image parallel to the character .yaml, overwriting the original, and potentialy larger, image.
If this was the intended behavior, I would have assumed the down scaling would be performed when an image is initially imported.
https://github.com/oobabooga/text-generation-webui/blob/cc8c7ed2093cbc747e7032420eae14b5b3c30311/modules/chat.py#L721
Suggested sollution:
Is there an existing issue for this?
Reproduction
alt.
Screenshot
No response
Logs
System Info