ComfyUI v0.38.0 adds Ming Image 0.1, MiniMax-H3 ControlNet Union 2.0, ID-V2V
Gremlation · reddit · 2026-09-30
ComfyUI v0.38.0 ships new open-source model support: Ming Image 0.1 (6B text-to-image for text-heavy design with transparent RGBA output), MiniMax-H3 Fun ControlNet Union 2.0, ID-V2V identity-preserving video-to-video on Wan 2.1/VACE, plus Qwen-Image 2.1 ControlNet and a tiny VAE. Node updates include LogC3/ACEScct color spaces, w6a8 quantization, Seedance 2.5 Draft mode, Seedream 5.0 Flash, and GPT-6 Sol/Luna tiers; Sora nodes removed. Performance work spans NPU async weight offload, CUDA graphs for ACE-Step 1.5, and non-blocking asset scans, with fixes for RGBA upscaling and the MiniMax-H3 VAE crash.
More from Multimodal
- Opus made a launch video for Violetto 1B in 50 minutes amid zero media coverage — tensorqt · 2026-09-30
- LoRA adapters break on distilled video models, long post explains why — burkov · 2026-09-30
- Creator builds temporally coherent AI depth-of-field on Marigold V2, fixing chronic flicker — AntonObukhov1 · 2026-09-30
- ByteDance's SplitMoE breaks the uniformity trap to scale video diffusion MoE models — ByteDance · 2026-09-30
- EmoRES: training-free vector steering boosts emotion TTS controllability by up to 118.8% — meta · 2026-09-30
- SoL-Refiner: one-step refinement turns low-res video into 4K with 8.91x latency speedup — Haozhe Liu · 2026-09-30