First short film made fully in ComfyUI with MiniMax H3, workflow shared
JL_Films · reddit · 2026-11-04
- Reddit user JLFilms shares The 13th, a short made entirely in ComfyUI with MiniMax H3.
- Stack: minimaxh3hybrid diffusion model + ref2v turbo 8-step LoRA at 768p.
- Ran 8 steps locally on an RTX 5060 Ti 16GB; rented RunningHub GPUs for 20-step runs.
- The Model Preview Override node is a must — previews now include audio for checking dialogue.
- Post-production in DaVinci Resolve 21, using a workflow modified from Pixaroma's.
- Tried Siingularity, but it alters the reference image and prompt adherence lags the base/hybrid models.
More from Multimodal
- Adobe Research & KAIST unveil MoviEdit for consistent multi-shot long-video editing — _akhaliq · 2026-11-04
- Midjourney SREF share: hard-edged color shapes and deep teal shadows — tisch_eins · 2026-11-04
- Voxel pagoda garden from an older Fable 5.5 checkpoint is the best attempt yet — legit_api · 2026-11-04
- CinemaWorld at UIST 2026: generative AR with LLMs and 3D scene generation for movies — ryosuzk · 2026-11-04
- HYPER3D Rigging debuts: prompt-editable auto-rigging for 3D characters with 250+ motion presets — _akhaliq · 2026-11-04
- ComfyUI-StudioLight v1.0: open-source studio lighting control for AI portraits with key, fill and rim lights — Neat_Two_8536 · 2026-11-04