Kandinsky open-sources S2V 1.0: speech-driven full-body avatar video in 2B and 21B sizes
Crazy-Repeat-2006 · reddit · 2026-11-05
Kandinsky released Kandinsky S2V 1.0, a speech-to-video avatar model: given a reference frame and a speech track, it animates the entire figure — not just the head — to match the speech. Audio acts as a condition; the model doesn't generate sound.
Open weights are on Hugging Face in two tiers: Lite 2B (plus a distilled Lite-distill variant) and Pro 21.4B. Project page: kandinskyavatar.github.io.
More from Multimodal
- Viggle Remake launches with open weights: recast any viral clip with your AI character at $0.96/10s — ZabihullahAtal · 2026-11-05
- Dev builds a handwriting engine that mimics penmanship styles from 1522 to today — floguo · 2026-11-05
- RefMods workflow keeps characters and voices consistent in MiniMax H3 on a 6 GB GPU — EzkaProductions · 2026-11-05
- Full WoW-parody music video made with Wan 2.2 in ComfyUI, music by Suno — El_Perro_Gordo · 2026-11-05
- codevember Day 3: a melting three.js demo titled 'Melt' — thespite · 2026-11-05
- HunyuanVideo 1.5 T2V FP8 outputs noisy images on AMD RX 7900 XTX — Fragrant-Throat7048 · 2026-11-05