MiniMax RefMod Tutorial: Reusable Character Identities Across Image, Video and Audio Without Training
Hearmeman98 · reddit · 2026-09-12
A community tutorial shares complete workflows for MiniMax RefMod that create reusable character/identity references — for image, video, and audio alike — without any model training.
The author provides downloadable workflows via Google Drive, covers how to prepare training images in the walkthrough, and offers a Runpod template with all workflows, models, and custom nodes preloaded for one-click reproduction.
Related event: Tutorials Emerge for Training-Free MiniMax RefMod Workflows in ComfyUI(4 posts)→
More from Multimodal
- Primal VAE released: 80M-param 2x VAE patches to match FLUX 1/2 latents — ostrisai · 2026-10-31
- One-shot 7-minute Bitfinex true-crime documentary made entirely by AI agents — toolstelegraph · 2026-10-31
- Distilling Krea2 into Anima with a LoRA trained on just 1,200 captions — Honest_Concert_6473 · 2026-10-31
- Open-source tool uses Qwen as a video judge to fix problem video segments — Either-Style3306 · 2026-10-31
- AI-made 17-minute dark comedy short debuts for Halloween, built on Voyager, ElevenLabs, Suno — ravisparikh · 2026-10-31
- A three.js Metahuman alternative with clothing exists, and it's getting better — jasonkneen · 2026-10-31