LAION Releases Humaneness Voice Converter for Local Use
The complete speech-to-speech checkpoint runs locally, but LAION’s own evaluation finds gains in predicted quality alongside weaker target-voice similarity.
Listen
AI narration
12:07
0:00 / 12:07
AI SummaryGenerated from this article
LAION released Humaneness Voice Converter, a speech-to-speech model that reproduces spoken content in a different speaker's voice while running entirely locally. The complete inference checkpoint includes the Chatterbox speech-token extractor, speaker encoder, flow decoder, and HiFT vocoder, requiring only a CUDA-enabled PyTorch installation.
LAION's evaluation shows trade-offs: the converter improves predicted content enjoyment and production quality compared to untouched Chatterbox but scores lower on target-speaker similarity. These metrics come from Humaneness Ears Medium, the same scoring system used during training, rather than independent human listening studies. The model card explicitly notes the absence of human mean opinion scores or word error rate results, and users should validate converted audio for missing words, altered emotion, and identity drift before deployment.
LAION’s Humaneness Voice Converter takes an existing spoken recording and attempts to reproduce its content in another consenting speaker’s voice, while retaining much of the original delivery. The release includes a complete inference checkpoint and a local command-line workflow. Users do not have to merge an adapter into another model.
LAION’s evaluation shows a trade-off. Compared with untouched Chatterbox, the converter scores better on predicted content enjoyment, predicted production quality, and source-emotion matching. It scores worse on target-speaker similarity. A rendition that a scoring model prefers may not sound most like the intended speaker, and neither result establishes that the words survived unchanged.
The Startrise release tracker lists the release on October 11, 2026. LAION’s model card supplies the substantive evidence: the runnable package, its evaluation conditions, and limitations that should accompany any local experiment.
A Complete Speech-to-Speech Checkpoint, Not a TTS Model
Humaneness Voice Converter is based on Resemble AI’s Chatterbox, with the voice-conversion architecture and training developed by LAION. It is an audio-to-audio system, not an LLM or a text-to-speech model, and is distinct from Humaneness Voice Small semantic TTS.
The source recording supplies speech content, timing, and performance. The target recording supplies the voice reference, documented as a clean, single-speaker clip lasting five to ten seconds.
The package includes the Chatterbox S3 speech-token extractor, CAMPPlus speaker encoder, score-conditioned flow decoder, HiFT vocoder, and frozen Humaneness Ears Medium score predictor. Users do not need to locate and combine several LoRA adapters before inference.
Chatterbox extracts roughly 25 semantic speech tokens per second from the source audio. The target supplies both a speaker embedding and a short acoustic prefix. Humaneness Ears Medium adds 99 conditioning scores, with time-varying emotion and style information drawn from the source and selected relatively stable voice attributes drawn from the target.
The decoder generates mel features, which represent the audio’s spectral content, and HiFT renders a 24 kHz waveform. Both target-conditioning inputs matter; this is not voice cloning from a text description alone.
How to Run Humaneness Voice Converter Locally
LAION’s local setup instructions assume a CUDA-enabled PyTorch installation. The reported test environment used Python 3.13.5, PyTorch 2.9.1+cu128, Chatterbox 0.1.7, and Transformers 5.14.1.
Work with Zeniteq
Let’s work together
We’re open to thoughtful collaborations with teams building in AI. Explore the ways we can work together.
The commands were validated inside an existing Jupiter environment that overrides several upstream Chatterbox dependency pins. LAION does not present that arrangement as a fresh, cross-platform dependency-resolution test. The version list documents a working environment; it does not guarantee that installation will succeed unchanged on every machine.
With platform-compatible CUDA PyTorch installed and the hf command available, download the repository first so its requirements file is accessible:
The --no-deps installation follows LAION’s documented arrangement. It does not resolve dependency conflicts automatically, so checking the resulting package versions is part of setup.
For one conversion, provide the source speech and an authorized target recording:
The validated input path accepts finite audio samples lasting one to 30 seconds, either mono or mixed down to mono. WAV works; other formats depend on the installed SoundFile build. For results comparable to the documented setup, use a clean five-to-ten-second target with one speaker.
LAION reports running the command on a GPU with a 6.528-second source and a 5.448-second target, producing a 6.52-second, 24 kHz WAV. This establishes a concrete tested path, without demonstrating performance across all recording conditions.
A fixed seed makes sampling repeatable on a fixed software and hardware stack. Different seeds produce different takes. Optional candidate selection can generate and rank several:
Candidate selection is a more expensive inference strategy and is not the default. With one candidate, the script does not run candidate ranking. The ranker combines standardized predicted enjoyment, target-speaker cosine similarity, and negative emotion error with equal weights. Its judgment remains an automated proxy, not a listener’s assessment.
Predicted Quality Improves While Speaker Similarity Falls
LAION’s reported comparison uses 40 curated source recordings and two fixed target voices. The default results are single-sample outputs; best-of-16 results add candidate selection.
Decoder
Predicted Content Enjoyment ↑
Predicted Production Quality ↑
Emotion MSE ↓
Target-Speaker Cosine ↑
Untouched Chatterbox S3Gen
6.169
7.766
0.113
0.821
Humaneness Voice Converter
6.453
8.009
0.092
0.773
Humaneness Voice Converter, best-of-16
6.460
8.009
0.072
0.795
The default converter gains 0.284 points in predicted content enjoyment and 0.243 in predicted production quality. Emotion mean squared error falls from 0.113 to 0.092, indicating closer matching on the source’s two strongest emotion axes under this evaluator.
Target-speaker cosine similarity falls from 0.821 to 0.773. Best-of-16 selection recovers some of that loss, reaching 0.795, but remains below untouched Chatterbox’s 0.821. Predicted enjoyment barely changes with selection, and the reported production-quality score stays the same. Candidate selection therefore looks more plausible as a tool for improving emotion matching and speaker consistency than for substantially raising these quality scores.
Cosine similarity measures similarity between speaker embeddings; it is not a percentage of voice-cloning accuracy. A lower score is relevant evidence about target matching, but it cannot identify every audible identity difference or verify who is speaking.
The Evaluation Is Not an Independent Listening Study
Content enjoyment, production quality, and emotion scores come from Humaneness Ears Medium, the same judge family used during training. These quality figures need careful reading.
LAION’s later full-flow training weighted predicted content enjoyment at 40%, source-emotion matching at 30%, and target-speaker similarity at 30%. Those weights help explain the model’s objectives, but they do not establish that listeners prefer the resulting compromise. Evaluating with a related training judge is not automatically invalid; improvements should be read as progress against that scoring system until independent evidence establishes broader benefits.
Sample selection also limits the comparison. The sources were curated using input quality, and ten source identifiers overlap the known reinforcement-learning source list. Speaker-level separation has not been proven. Two target voices offer a controlled comparison, but little evidence about how consistently the trade-off holds across different speakers.
A separate 93-source comparison against LAION’s 24k predecessor shows uneven progress. The 40k version improved predicted enjoyment from 5.930 to 6.074 and production quality from 7.790 to 7.876, while emotion error worsened from 0.289 to 0.320. That set excludes training source identifiers but is not proven speaker-disjoint.
The model card explicitly states that there are no independent human mean opinion score (MOS) or word error rate (WER) results for this release. The demo is also not a blinded listening study.
The published scores establish neither independent human preference nor transcript fidelity. Local evaluation should compare the converted audio with the source for missing or altered words, unexpected emotion, audible artifacts, and identity drift. English and German appear in the training and evaluation material; informal processing of other languages is not a multilingual quality guarantee.
Open Licensing Does Not Replace Voice Consent
LAION publishes its additions and documentation under CC BY 4.0. Bundled Chatterbox-derived components retain their upstream MIT license and attribution, while external demo recordings retain their own rights.
The checkpoint is accessible for experimentation, but its licensing does not grant permission to use somebody’s voice. Model permissions and recording or speaker rights are separate questions.
LAION identifies research, comparative evaluation, and authorized dataset augmentation as intended uses. Its guidance calls for informed consent where appropriate, disclosure of converted speech, and no impersonation, fraud, deception, or identity spoofing. The card distinguishes those ethical recommendations from additional legal restrictions on the CC BY 4.0 license.
For a practical experiment, establish permission for both recordings and the intended output before conversion. Consent to provide a sample should not be treated as consent to every downstream use. If converted speech will be distributed, decide how its synthetic transformation will be disclosed and whether the target speaker should approve it.
Dataset augmentation needs the same care. A convincing output can still change content or emotion, so synthetic examples need validation before they become training or evaluation material. Downstream deployment also requires reviewing applicable voice rights, dataset terms, jurisdictional law, and safety controls.
The strongest reason to examine this release is its inspectability: LAION provides a complete local inference package and reports a trade-off instead of presenting an unqualified quality win. It supports a reproducible experiment, not a verified replacement for a particular speaker. For identity-sensitive use, untouched Chatterbox remains an important baseline, and independent listening and content-fidelity evidence should carry more weight than the internal proxy gains.
