MultiTalk generates multi-person lip-synced conversation videos from audio, hits NeurIPS 2025
solyarisoftware · x · 2026-10-24
Open-source MultiTalk (NeurIPS 2025, 3k GitHub stars) generates multi-person conversational videos from multi-track audio plus a reference image, with lip-sync and prompt-controlled interactions, also supporting cartoons and singing. It runs 480p on a single RTX 4090, the community Wan2GP build needs only 8GB VRAM, and it integrates with ComfyUI and Colab. The team followed up with InfiniteTalk and LongCat-Video-Avatar, with v1.5 compressing inference to 8 steps.
More from Multimodal
- Just Eat's AI food photos turned tiger prawns into baby tigers — a linguistics lesson in why — aakashgupta · 2026-10-24
- Community tips for MOSS: Turbo Lora settings, samplers, and SLG tuning — mikemend · 2026-10-24
- Dreamy AI-generated nebula-eared creature makes the rounds among AI artists — LudovicCreator · 2026-10-24
- LoRA trained on 39 images in 40 minutes makes MOSS-Image output real transparent PNGs — gatortux · 2026-10-24
- Dev uses Sol 6.1 to generate game terrain and props, then arranges them in his editor — Dimillian · 2026-10-24
- Seedance 2.5 generates 30-second consistent AI influencer dance videos — aftahi_ai · 2026-10-24