ModelÜberprüftExtern

Chrome On Aituber Irodori Assets

fp16 ONNX export of Irodori-TTS, packaged for in-browser WebGPU inference with chrome-on-aituber. The browser app downloads these files directly (CORS) and stores them in browser storage; nothing is sent to a server.

Quellehuggingface.co/shinshin86/chrome-on-aituber-irodori-assets/tree/cd80d15900ddfc057fb7bdbb86c6e17e59c2952c

ArtAudio modelVersionRevision cd80d15Verleger@shinshin86CBewertung

Modellquelle

Art: Audio model, Text encoder, Text model
Dateigröße: 223 KB–2,79 MB
Version: vRevision cd80d15
Aktualisiert: 3. Okt. 2026
Verleger: @shinshin86
Downloads: 0
Parameter: 150M
Aufgaben: Audio
Quelle: Hugging Face

Quellenbeschreibung

Irodori TTS WebGPU assets for chrome-on-aituber

Path	Description
`manifest.json`	Download manifest (version, file list, sizes)
`onnx_fp16/`	fp16 ONNX graphs + external weights (text/speaker/duration/DiT/DACVAE)
`tokenizer/llmjp_tok/`	llm-jp-3-150m tokenizer files
`runtime/`	Bundled WebGPU runtime (`pipeline.mjs`) and ONNX Runtime wasm loaders
`licenses/`	Full license texts and third-party notices for everything above

The ONNX export and the inference runtime are built with ngc-shj/irodori-tts-webgpu. The reproducible build recipe lives in the chrome-on-aituber repository.

Licenses and attribution

This repository redistributes derivative weights and runtime code. See licenses/ for the full texts and licenses/THIRD_PARTY_NOTICES.txt / licenses/NOTICE for details.

Aratako/Irodori-TTS-500M-v3 — MIT
Aratako/Semantic-DACVAE-Japanese-32dim — MIT (derived from facebook/dacvae-watermarked, Apache-2.0)
llm-jp/llm-jp-3-150m tokenizer — Apache-2.0
ngc-shj/irodori-tts-webgpu — MIT
onnxruntime-web — MIT
@huggingface/tokenizers — Apache-2.0

Responsible use

This model performs reference-audio-based voice synthesis (voice cloning). Do not use it for impersonation, fraud, or generating misleading content. Only use reference audio you have the right to use. Note that, unlike the Python reference implementation of Irodori-TTS, this browser runtime does not apply SilentCipher watermarking to generated audio (see licenses/THIRD_PARTY_NOTICES.txt).

Quellen

1 Quelle

Hugging FacePrimärVersion cd80d15900ddfc057fb7bdbb86c6e17e59c2952c

huggingface.co/shinshin86/chrome-on-aituber-irodori-assets/tree/cd80d15900ddfc057fb7bdbb86c6e17e59c2952c

Verifiziert 3. Okt.

Modellartefakte

6 Artefakte

onnx_fp16/dacvae_decoder.onnx

onnx · 826 KB · SHA-256 1dd2e67a0a83…8c03 · Hugging Face

Herunterladen

onnx_fp16/dacvae_encoder.onnx

onnx · 883 KB · SHA-256 d7abab239ea9…7a51 · Hugging Face

Herunterladen

onnx_fp16/dit.onnx

onnx · 2,79 MB · SHA-256 b4dbc4ad12a4…4ec4 · Hugging Face

Herunterladen

onnx_fp16/duration.onnx

onnx · 223 KB · SHA-256 365e08ef08b7…f214 · Hugging Face

Herunterladen

onnx_fp16/speaker_encoder.onnx

onnx · 1,48 MB · SHA-256 95bcccfa47b2…57aa · Hugging Face

Herunterladen

onnx_fp16/text_encoder.onnx

onnx · 1,68 MB · SHA-256 d4dda1ced388…b2ef · Hugging Face

Herunterladen

Quellenauszüge

3 Auszüge

Readme11. Julihuggingface.co/shinshin86/chrome-on-aituber-irodori-assets

--- license: mit language: - ja pipeline_tag: text-to-speech tags: - tts - onnx - webgpu - irodori-tts --- # Irodori TTS WebGPU assets for chrome-on-aituber fp16 ONNX export of [Irodori-TTS](https://github.com/Aratako/Irodori-TTS), packaged for in-browser WebGPU inference with [chrome-on-aituber](https://github.com/shinshin86/chrome-on-aituber). The browser app downloads these files directly (CORS) and stores them in browser storage; nothing is sent to a server. ## Contents | Path | Description | | --- | --- | | `manifest.json` | Download manifest (version, file list, sizes) | | `onnx_fp16/` | fp16 ONNX graphs + external weights (text/speaker/duration/DiT/DACVAE) | | `tokenizer/llmjp_tok/` | llm-jp-3-150m tokenizer files | | `runtime/` | Bundled WebGPU runtime (`pipeline.mjs`) and ONNX Runtime wasm loaders | | `licenses/` | Full license texts and third-party notices for everything above | The ONNX export and the inference runtime are built with [ngc-shj/irodori-tts-webgpu](https://github.com/ngc-shj/irodori-tts-webgpu). The reproducible build recipe lives in the [chrome-on-aituber repository](https://github.com/shinshin86/chrome-on-aituber/tree/main/scripts/irodori). ## Licenses and attribution This repository redistributes derivative weights and runtime code. See `licenses/` for the full texts and `licenses/THIRD_PARTY_NOTICES.txt` / `licenses/NOTICE` for details. - [Aratako/Irodori-TTS-500M-v3](https://huggingface.co/Aratako/Irodori-TTS-500M-v3) — MIT - [Aratako/Semantic-DACVAE-Japanese-32dim](https://huggingface.co/Aratako/Semantic-DACVAE-Japanese-32dim) — MIT (derived from [facebook/dacvae-watermarked](https://huggingface.co/facebook/dacvae-watermarked), Apache-2.0) - [llm-jp/llm-jp-3-150m](https://huggingface.co/llm-jp/llm-jp-3-150m) tokenizer — Apache-2.0 - [ngc-shj/irodori-tts-webgpu](https://github.com/ngc-shj/irodori-tts-webgpu) — MIT - [onnxruntime-web](https://github.com/microsoft/onnxruntime) — MIT - [@huggingface/tokenizers](https://github.com/huggingface/tokenizers.js) — Apache-2.0 ## Responsible use This model performs reference-audio-based voice synthesis (voice cloning). Do not use it for impersonation, fraud, or generating misleading content. Only use reference audio you have the right to use. Note that, unlike the Python reference implementation of Irodori-TTS, this browser runtime does not apply SilentCipher watermarking to g...

Note11. Julihuggingface.co/shinshin86/chrome-on-aituber-irodori-assets

Source context: 0 downloads · 0 likes · Pipeline text-to-speech · Repo shinshin86/chrome-on-aituber-irodori-assets

Note13. Junihuggingface.co/api/models/shinshin86/chrome-on-aituber-irodori-assets

Source context: 0 downloads · 0 likes · Pipeline text-to-speech · Repo shinshin86/chrome-on-aituber-irodori-assets