Refactor of sd_handler.py

[w4ffl35]

Refactor of SD Handler

This is a refactor of the stable diffusion handler class.

This will provide a speed boost to loading and using stable diffusion overall, as well as the canvas when images are being generated. This improves the "real-time" drawing performance of the canvas and image generation.

Logger from info to debug

All logger info calls have been moved to debug calls. Several workers have been improved, as well has the handling of exceptions and closing of threads.

Fixes prompt embed loading

There has been a bug in the code for several weeks (or months) which related to loading prompt embeds - it is patched in this PR.

Model loading fixes and more

• Fixes loading between LLM and Stable diffusion
• Fixes mute / unmute microphone
• Fixes text-to-speech => text-to-image pipeline
• Adds support for SPD text-to-speech generation
• Improves performance of threads and more

---

Video preview here

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues found.

Scanned Manifest Files

setup.py

• transformers@4.38.2
• transformers@4.38.1