Lipsync AI music video with Suno and MiniMax in ComfyUI, rendered on RTX 5070Ti
xXleekXx · reddit · 2026-10-29
A creator shares a full pipeline for making lipsync music videos with Suno music and MiniMax in ComfyUI, rendered locally on an RTX 5070Ti — each full video takes about 2 hours to generate:
- Built on a public MiniMax lipsync workflow, with lots of trial-and-error assisted by Claude, which also writes each clip's prompt
- You need: the full song split into vocals/instrumental, timestamps for each lyric line, and at least two character reference images
- Usually 2-3 passes or more; the preview stream lets you catch desync, weird shots and bad angles mid-generation and stop early
More MVs made with the same method are on the author's YouTube channel.
More from Multimodal
- Skip ComfyUI: run Flux-3-Image-Klein locally with stable-diffusion.cpp in four files — srigi · 2026-10-29
- Dev uses Opus 5.5 as creative partner for creature creator: heads and hands come out 'massively good' — repligate · 2026-10-29
- Terraformer turns your city design into a dressed Unreal Engine 5 level using only open weights — ki11ogs · 2026-10-29
- Top 20 Midjourney light-condition prompts, from golden hour to moonlight — michaelrabone · 2026-10-29
- A Midjourney SREF code for 1950s sci-fi vibes, plus a 500+ code style library — tisch_eins · 2026-10-29
- Inaugural Lumara AI Film Festival wraps in NYC with sold-out crowds for AI-assisted cinema — Kyrannio · 2026-10-29