Minimax Video Face-Swap Fails Past 5s? Prompting & Param Tips
Mediocre-Toe3212 · reddit · 2026-08-06
While running Minimax video generation in ComfyUI for reference-based face swaps, the author encountered a matching failure: the model loses facial lock when generation exceeds 5 seconds.
Through extensive testing, several workarounds were identified:
- Step Count: 12-15 steps work best. The default 20 steps cause the model to overthink and fail the match.
- Ref Image Size: Setting to MAX yields better results than match.
- Prompting: Describing specific actions (e.g., "attacking with knives") tends to alter the original video scene. Sticking to subject and static environment descriptions (e.g., "surrounded by men with knives") improves success rates.
The author still cannot reliably generate swaps beyond 5-6 seconds and is seeking further community advice.
Related event: MiniMax H3 Local Tests: Fast but Flawed in Quality and Consistency(19 posts)→
More from Multimodal
- Kling 4.0 fully available, shows markedly better motion coherence in fast-action scenes — alifcoder · 2026-10-16
- MJev-Spatial: Single-Pass Classification Beats Autoregressive VLMs on Spatial Reasoning — mindlab-research · 2026-10-16
- H-DDCM: Hadamard Codebooks Make Diffusion-Based Image and Video Compression Scalable — Technion · 2026-10-16
- ComfyUI v0.41.0 ships with Flux3 image model and MiniMax-H3 LightVAE support — Fcking_Chuck · 2026-10-16
- AI-generated short film 'The Dragon Rider' glides cinematically above the clouds — jijo6686 · 2026-10-16
- Realistic candid Midjourney portrait prompt: natural skin, lived-in bedroom light — tisch_eins · 2026-10-16