Repurposing MiniMax H3 for Single-Image Editing: A ComfyUI Workflow Guide
Patient_Ratio4177 · reddit · 2026-08-14
A developer shared a complete ComfyUI workflow repurposing the video model MiniMax H3 for single-image editing.
- Core Components: Uses a hybrid checkpoint combining the strengths of FL2VA and Ref2VA, paired with a specific Video VAE for rendering single images to avoid blurriness and grid artifacts caused by simply extracting the first frame of a video.
- Engineering Details: Provides a patched nodesminimaxh3.py file to bypass ComfyUI's 5-frame minimum limit. Recommends the ThisIsFine LoRA for details and a Turbo LoRA for faster generation.
- Performance & Results: Takes about 8 seconds to generate a 1920×1088 image on an RTX 5090. The author notes it outperforms previous Krea 2 workflows in character fidelity, 3D scene handling, and composition, supporting outfit swaps, body type changes, stylization, and depth-map-based reposing.
More from Multimodal
- FrankenSim builds a physics-based virtual orchestra, piano and drum demos live — doodlestein · 2026-10-16
- vLLM officially supports Muse Spark Open: multimodal, 1M context, agentic — vllm_project · 2026-10-16
- InfoVis: 9B unified model tops open-source slide generation and editing, trained on 13M records — Jiahao Wang · 2026-10-16
- RepForcing: feature-space supervision improves few-step autoregressive video distillation — zju · 2026-10-16
- SCOPE: semantic span-aware GRPO objective improves T2I prompt rewriting — university-of-rochester · 2026-10-16
- Orbia benchmark asks: do generated videos form a consistent 3D world? — tencent · 2026-10-16