RefMods makes character LoRAs from a few images — plus near-perfect voice cloning
AlsterwasserHH · reddit · 2026-10-01
A Reddit user is blown away by RefMods in MH3: instead of training LoRAs (which he tried on Krea2), RefMods builds a character model from a handful of unlabeled images in seconds, producing a reusable safetensors file with stunning facial likeness and no distortion.
The kicker: drop an mp3 of voice audio into the same folder and it generates a near 1:1 voice clone inside the same safetensors file.
To reproduce: install the ComfyUI custom nodes and download the workflow; walkthrough video here.
More from Multimodal
- Solo-built home redesign app chains Opus 5.5, Gemini, SAM 3.1 and Ideogram for $1.50 per house — venturetwins · 2026-10-01
- Free Dot avatar maker: 25 shapes, 12 materials, 32 styles, no sign-up — yihui_indie · 2026-10-01
- SAM 3.1 + Ideogram 4.5 app restyles entire Zillow listings for $1.50 per house, $0.06 per edit — venturetwins · 2026-10-01
- Queue Workbench: Open-Source ComfyUI Plugin Fixes the Stock Queue's Blind Spots — Smegnigma · 2026-10-01
- Kling 4.0 promises 30s 4K clips, but the public beta is still 720p — lmoroney · 2026-10-01
- Full Recipe for the Viral Hotel Lobby Migos Video Swap: Reference Frames + Tagged Prompts — marupelkar · 2026-10-01