duyquangnvx/vieneu-web-models

vieneu-web models

The model files that vieneu-web loads in the browser. vieneu-web runs VieNeu-TTS v3 Turbo with onnxruntime-web, for speech and voice cloning. Load files from a pinned commit of this repo, never from main.

manifest.json lists every file with its size, sha256 and tier (text, speech, cloning), and records the pinned upstream sources. The files are built by tools/mirror in the vieneu-web repo.

Sources

Files	From
config.json, tokenizer.json, vieneu_*, speaker_encoder.onnx	pnnbao-ump/VieNeu-TTS-v3-Turbo at 5f2a3e9 (onnx_int8/, and the repo root for the speaker encoder)
audio_head_*.onnx, text_head_gather.onnx, heads.bin, heads.json	Generated from onnx_int8/vieneu_v3_heads.npz of the same repo
moss_audio_tokenizer_*, codec_browser_onnx_meta.json	OpenMOSS-Team/MOSS-Audio-Tokenizer-Nano-ONNX at ceff0d0
voices.json	The v3 Turbo Preset Voices of VieNeu-TTS at 04044e1, converted to vieneu-web's Voice File shape
sea_g2p.bin	The dictionary of sea-g2p 0.9.1, cut to Vietnamese and English

License

Apache 2.0, the license of every source above; see LICENSE. The weights, voices and dictionary belong to their upstream authors. This repo only redistributes them, some in converted form.