LAION releases open Humaneness Voice Converter that preserves emotion and timing
lmoroney · x · 2026-10-13
LAION published Humaneness Voice Converter on Hugging Face (October 11), a Chatterbox-based speech-to-speech model: give it a spoken source and a 5–10 second clip of a consenting target speaker, and it converts the content into the target voice while retaining much of the original timing, emphasis, and emotion.
Key points:
- Ships as a complete inference checkpoint: Chatterbox S3 speech-token extractor, CAMPPlus speaker encoder, score-conditioned flow decoder, HiFT vocoder, and the frozen Ears Medium score predictor; outputs 24 kHz audio in 10 flow steps.
- LAION's additions are CC BY 4.0; upstream Chatterbox components remain MIT.
- On a 40-source listening set, predicted Content Enjoyment rose from 6.169 to 6.453 and Production Quality from 7.766 to 8.009 versus untouched Chatterbox, with lower emotion error; target-speaker cosine similarity dropped slightly (0.821 → 0.773).
- No independent human MOS or WER yet; English and German are in training data, other languages have no quality guarantees.
- The model card mandates consent, disclosure of converted speech, and no impersonation.
Related event: LAION Open-Sources Humaneness Voice Converter(2 posts)→
More from Multimodal
- MiniCPM-V-4.7-1B, a lightweight on-device multimodal model, trends on Hugging Face — openbmb · 2026-10-13
- Eerie black-eyed AI portrait goes viral for its unsettling gaze — PierceLilholt · 2026-10-13
- AI recreates Metal Gear Solid PS1 cutscenes starring virtual idol Psyopqueen — tetsuoai · 2026-10-13
- Indie anime model MageTrail V0.4 shows artist-style learning from ~$20; 6k-artist run priced at $1-2k — Turbulent-Bass-649 · 2026-10-13
- Entire movie rendered as one sequence: FilmCraft shows cinematic AI long take — Super_Range45 · 2026-10-13
- Meta's Muse Wonder (Max) tested on one-shot 3D scenes and front-end vs top models — arena · 2026-10-13