Aller au contenu
ModelPrésentéExterne
Browse files

Enclave Tts

Voice-model assets used by the Enclave browser extension to synthesize speech locally, on-device. Hosted here only so the extension can fetch them once and cache them; nothing about how they're used is changed.

TypeOtherVersionFlow lm flow int8CClasse

Source du modèle

Type
Other
Taille du fichier
5,91–72,8 MB
Version
vFlow lm flow int8
Mis à jour
26 oct. 2026
Source
Hugging Face

Description de la source

Enclave TTS assets (pocket-tts, ONNX)

Voice-model assets used by the Enclave browser extension to synthesize speech locally, on-device. Hosted here only so the extension can fetch them once and cache them; nothing about how they're used is changed.

Attribution & license

These weights are Kyutai pocket-tts, licensed CC-BY-4.0. This repository redistributes them under the same license, with attribution to Kyutai.

Changes made to the original:

  • Exported from the official safetensors weights to ONNX and int8-quantized, using KevinAHM/pocket-tts-onnx-export (export code: MIT).
  • Only the alba voice is included (from kyutai/tts-voices → alba-mackenna, CC-BY-4.0). The other voices from the upstream bundle — some of which are CC-BY-NC — are not redistributed here.

Acceptable use

Per Kyutai's model terms: use must comply with applicable law and must not involve voice impersonation/cloning without consent, misinformation, or other harmful use. Enclave uses a single fixed preset voice to read the assistant's own generated replies aloud — no cloning.

Contents

bundle.json, tokenizer.model, bos_before_voice.npy, text_conditioner_int8.onnx, flow_lm_main_int8.onnx, flow_lm_flow_int8.onnx, mimi_decoder_int8.onnx, voices.bin (alba only).

Regenerate with node scripts/prepare-tts-assets.mjs in the Enclave repo.

Sources

1 source
Hugging FacePrincipalVersion 181fb77f64ef6196c06c185972880814c0a9bd8d
huggingface.co/team-edt/enclave-tts/tree/181fb77f64ef6196c06c185972880814c0a9bd8d

Vérifié 26 oct.

Download files

Artefacts du modèle

5 artefacts

flow_lm_flow_int8.onnx

onnx · 9,50 MB · SHA-256 3dd781ee5abe…635d · Hugging Face

Download file

flow_lm_main_int8.onnx

onnx · 72,8 MB · SHA-256 f9bd8106b79a…0116 · Hugging Face

Download file

mimi_decoder_int8.onnx

onnx · 21,6 MB · SHA-256 3630450a3297…f925 · Hugging Face

Download file

text_conditioner_int8.onnx

onnx · 15,6 MB · SHA-256 24c0642bff11…5ac3 · Hugging Face

Download file

voices.bin

bin · 5,91 MB · SHA-256 5c7e96387e61…da79 · Hugging Face

Download file

Extraits de sources

2 extraits

--- license: cc-by-4.0 tags: - text-to-speech - onnx - pocket-tts --- # Enclave TTS assets (pocket-tts, ONNX) Voice-model assets used by the [Enclave](https://github.com/abhishekray-edu/enclave) browser extension to synthesize speech **locally, on-device**. Hosted here only so the extension can fetch them once and cache them; nothing about how they're used is changed. ## Attribution & license These weights are **[Kyutai pocket-tts](https://huggingface.co/kyutai/pocket-tts)**, licensed **[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)**. This repository redistributes them under the same license, with attribution to Kyutai. **Changes made to the original:** - Exported from the official safetensors weights to **ONNX** and **int8-quantized**, using [KevinAHM/pocket-tts-onnx-export](https://github.com/KevinAHM/pocket-tts-onnx-export) (export code: MIT). - **Only the `alba` voice** is included (from [kyutai/tts-voices](https://huggingface.co/kyutai/tts-voices) → `alba-mackenna`, CC-BY-4.0). The other voices from the upstream bundle — some of which are CC-BY-**NC** — are **not** redistributed here. ## Acceptable use Per Kyutai's model terms: use must comply with applicable law and must not involve voice impersonation/cloning without consent, misinformation, or other harmful use. Enclave uses a single fixed preset voice to read the assistant's own generated replies aloud — no cloning. ## Contents `bundle.json`, `tokenizer.model`, `bos_before_voice.npy`, `text_conditioner_int8.onnx`, `flow_lm_main_int8.onnx`, `flow_lm_flow_int8.onnx`, `mimi_decoder_int8.onnx`, `voices.bin` (alba only). Regenerate with `node scripts/prepare-tts-assets.mjs` in the Enclave repo.

Source context: 0 downloads · 0 likes · Pipeline text-to-speech · Library pocket-tts · Repo team-edt/enclave-tts