Skip to content

DiffusionBear v0.4.5 beta

Choose a tag to compare

@OuincheWinch OuincheWinch released this 08 Nov 19:57

[0.4.5] — beta — 2026-11-08

Package: DiffusionBear-0.4.5-arm64.zip — 384 MB
SHA-256: 196afeb2c43c8622ca13b9779e4c76ee619a333f475c1f3a678d863b041cb330

DiffusionBear v0.4.5

🎥 Video Tutorial

Watch the walkthrough: DiffusionBear video tutorial

✨ What's New

  • Native MLX SDXL engine — rewritten from scratch: removed the second venv-sdxl runtime and mlx_diffuser; we now own the MLX UNet, dual CLIP encoders, VAE decoder, schedulers, LoRA stack, 4/8-bit quantization, resident text encoders and TAESD. The second runtime removal cut the app bundle from 1.7 GB to 1.3 GB and the release zip from 510 MB to 385 MB. 44-image benchmark: Juggernaut XL Lightning 9.4s, Juggernaut XI 40.8s, RealVisXL V5 40.7s, Pony 106.9s; 0 errors.
  • Long prompts and mandatory adapters: SDXL/Pony prompts are split across 77-token CLIP windows instead of silently truncated; Qwen DMD2 and MOSS DMD adapters auto-restore when missing; Qwen 2.1 reference images now warn to use Qwen-Image Edit for instruction edits.
  • Qwen base self-healing: the base pack now supplies its own tokenizer/scheduler/model_index instead of borrowing another Qwen install; a bad processor/tokenizer symlink (for example the MOSS qwen3_5-te-q8g64 text encoder) is detected and replaced with Qwen-Image's own processor. Warm Qwen pipelines are invalidated when model files change and sanitized after every request; incomplete component weights fail with a clear re-download error instead of rendering a texture.
  • Licence display and validation reviewed.
  • LoRA management polished, including the Krea 2 Turbo LoRA recipe.
  • MOSS-Image batching fixed: the second image in a cached batch no longer fails during VAE decode.
  • MOSS LoRAs fixed: detected/registered as MOSS, selectable in the architecture dropdown, and the real tensor layout now overrides Civitai metadata.
  • Pony Diffusion V6 XL: corrected to a working Diffusers download source.
  • Z-Image / FLUX samplers: Euler now maps to the flow-match scheduler instead of the Krea 2 scheduler.
  • French model sizes: shown as GB instead of Go.
  • Qwen-Image integration refined.
  • Minor bug fixes.

🍎 First Launch on macOS (Gatekeeper)

Before opening the app for the first time, move DiffusionBear.app to /Applications and run:
xattr -cr /Applications/DiffusionBear.app