ComfyUI v0.44.0 ships MiniMax H3 editing nodes, GPT-6.1 models and ProRes export
Gremlation · reddit · 2026-11-05
ComfyUI v0.44.0 is out with a large update:
Partner node updates
- MiniMax H3 Max editing nodes: Recast, Extend, Reframe, Insert, Lip Sync and Camera Control
- Pruna P-Video Upscale, Topaz Wonder 4, Sonilo Music 1.5 (lyrics-to-song), Ideogram 4.5 Precise Edit context windows, and OpenAI GPT-6.1 Terra and Sol models
- Flux Video Upscale V2 targets 1080p/2K/4K; Seedance 1.x nodes deprecated
Open-source model support
- P-H3-Image: Pruna AI's few-step image model built on MiniMax H3
- MiniMax H3 latent upscaler, Sol-H3-VAE, and longer H3 video via overlapping context windows
New nodes & improvements
- Text Generate V2 with growing image inputs; separate VAE compute dtype
- ProRes video saving with chroma subsampling; fp8e4m3fnfast no longer falls back to bf16 under dynamic VRAM
More from Multimodal
- GenAI artist serves as art director on official music video "Proteggi il Cielo" — azed_ai · 2026-11-05
- Glowing bottle beach scene made with Seedance 2.5 for Luma's creative prompt challenge — LudovicCreator · 2026-11-05
- AI turns Marilyn, Freddie Mercury and Elvis into retro alien celebrity posters — azed_ai · 2026-11-05
- ElevenLabs ships Eleven v4 Turbo with ~150ms latency, REST, streaming and SDKs — azed_ai · 2026-11-05
- LFM3-Omni-2B ported to run fully offline on iPhone at ~80 tok/s with custom WebGPU kernels — JosephJacks_ · 2026-11-05
- Dot's creative ChatGPT image-gen workflow draws praise: storybook art beyond one-shot prompts — thederbiedone · 2026-11-05