New ComfyUI Stills mode for MiniMax H3 via custom Latent & VAE Decode node
shootthesound · reddit · 2026-09-26
After days of digging into MiniMax H3 latents and VAE decoding, the author open-sourced a new Stills mode for ComfyUI (ComfyUI-Fizgig-H3-Still) with a custom Latent & VAE Decode node.
The author reports a major quality improvement over the default pipeline, though not perfect. It's fast: the provided workflow uses the turbo LoRA only at a low setting, which can be disabled in favor of more steps. Results depend heavily on the specific turbo LoRA and sampler/scheduler combo—the author hunted down one strong pairing but suspects even better combinations exist.
More from Multimodal
- Adobe MAX: MCP across Creative Cloud, AI agents editing layers, Firefly Builder and Graph — umesh_ai · 2026-11-11
- Comment: China leads the AI video revolution while the West argued about AI art — TinfoilTricorn · 2026-11-11
- "QT your Geisha" meme takes over X with stunning porcelain-style AI portraits — LudovicCreator · 2026-11-11
- MiniMax H3-Image early access test: 4 images, one coherent cyberpunk world — Aiden_Tech_Ai · 2026-11-11
- MiniMax H3 Image generates a GTA6-style hurricane scene from a single prompt — socialwithaayan · 2026-11-11
- Dragon Ball reimagined as a 1971 movie: another hit AI retro trailer — techhalla · 2026-11-11