ComfyUI v0.36.0 Adds FastH3 Video, Marigold V2 Estimation and YuE2 Music Support
Gremlation · reddit · 2026-09-19
ComfyUI released v0.36.0 with a batch of notable updates:
New model support
- FastVideo FastH3: 8-step distilled MiniMax H3 for text/image-to-video with native audio
- Marigold V2: single-image depth, surface normal, and albedo estimation
- YuE2: music generation from style prompts and lyrics, AMD fixes, songs up to 900 seconds
Node updates
- Video Concatenate for joining clips; Generic Loops for iterating lists in graphs; Gemini V3 node (V2 deprecated)
- Partner nodes: Tripo P2 (text/image/multiview-to-3D), Pruna P-Video-2 (720p/1080p + draft mode), Flux Video Edit (text-instruction video editing), Bria edit suite
Performance & fixes
- Auto-detect --fast-disk; lower MiniMax-H3 VAE memory; fast llama rope kernels; AMD Windows VA quota raised to 4TB; fixed Add Noise alpha and H3 Fun ControlNet bugs
New workflow templates cover FastH3, Marigold V2, and YuE2.
More from Multimodal
- Experimental video mixes visual styles, generated with Seedance 2.5 on Adobe Firefly — umesh_ai · 2026-11-04
- Hailuo AI claims H3 scored perfect on language tests; Japanese creator praises its Japanese signage fidelity — Hailuo_AI · 2026-11-04
- YANAMI: 400K-dialog multimodal dataset shows models still fail at stateful multiturn interaction — Shengyu Wang · 2026-11-04
- Open-source Idol model turns a single image into 3D in 45s on 8GB VRAM — Promptmethus · 2026-11-04
- ComfyUI v0.44.0 released with new Patina texture nodes — Fcking_Chuck · 2026-11-04
- Krea 3 technical report released: a deep dive into data, architecture and training — bdsqlsz · 2026-11-04