cstr/melotts-en-v2-GGUF

MeloTTS English V2 — GGUF

myshell-ai/MeloTTS English V2 checkpoint converted to GGUF for CrispASR.

Model Details

• Architecture: VITS2 (52M params)
• Output: 44.1 kHz mono PCM
• Speakers: EN-US (0), EN-BR (1), EN-INDIA (2), EN-AU (4)
• Format: F16 GGUF (97 MB)
• License: MIT

Usage

# Auto-download + synthesize
crispasr --backend melotts -m auto --tts "Hello world." --tts-output hello.wav

# Manual download
wget https://huggingface.co/cstr/melotts-en-v2-GGUF/resolve/main/melotts-en-v2-f16.gguf
crispasr -m melotts-en-v2-f16.gguf --tts "Hello world." --tts-output hello.wav

Conversion

Converted from the official PyTorch checkpoint using:

python models/convert-melotts-to-gguf.py     --ckpt checkpoint.pth --config config.json     --output melotts-en-v2-f16.gguf

Weight-norm pairs fused. Embeddings and SDP/DP weights stored as F32 for precision; everything else F16.

Provenance and EU AI Act Art. 53 note

• Upstream model: myshell-ai/MeloTTS.
• Upstream licence: mit. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• What was done here: format conversion and/or quantisation only (GGUF). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository.
• Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.