ComfyUI Tutorial: MiniMax H3 Face Swap Workflow Runs on 6GB VRAM
cgpixel23 · reddit · 2026-09-03
A custom MiniMax H3 Ref2Vid ComfyUI workflow combines SAM3 masking with face swapping: load a reference face plus source video, define the mask with a simple prompt like face or head, and generate the swapped video. Optimizations (Low VRAM Attention, Chunk FeedForward, SLA Attention, Sol-Attn, Spectrum, INT8Conv) reportedly bring it down to 6GB VRAM. Full tutorial on Civitai and YouTube.
Related event: ComfyUI Tutorial Runs MiniMax H3 Face-Swap Workflows on 6GB VRAM(2 posts)→
More from Multimodal
- Ant open-sources 5 Ming-3.0 speech models spanning TTS, audio understanding and ASR — Med1_Ai · 2026-10-21
- What keeps AI videos watchable: hooks, consistency and a payoff — aftahi_ai · 2026-10-21
- MORPHJUTSU skill in Flova transfers any video's motion to your character — aftahi_ai · 2026-10-21
- NVIDIA open-sources MotionBricks: 350K motion skills at 15,000 FPS, 2ms latency — aftahi_ai · 2026-10-21
- Runway joins Claude Marketplace, letting enterprises spend Anthropic budget on video generation — runwayml · 2026-10-21
- Kokon 0.10 turns Qwen-Image 2.1's text encoder into a prompt writer, no extra model — Ok-Director4174 · 2026-10-21