Skip to content
ModelListedExternal
Browse files

Roo Voice NeuTTS GGUF

Single-voice identity-locked fine-tune of NeuTTS-Air (Qwen2.5-0.5B backbone) — the custom voice model behind the Roo Voice desktop app.

Sourcehuggingface.co/abliter8-ai/Roo-Voice-NeuTTS-GGUF/tree/4b22111ce98b3821767e699e03c03d83b7c1b754
KindOtherBase modelneuphonic/neutts-airVersionRevision 4b22111Bgrade

Model source

Kind
Other
File size
441–568 MB
Base model
neuphonic/neutts-air
Version
vRevision 4b22111
Updated
Oct 19, 2026
Downloads
202
Source
Hugging Face

Source description

Roo-Voice-NeuTTS-GGUF

Single-voice identity-locked fine-tune of NeuTTS-Air (Qwen2.5-0.5B backbone) — the custom voice model behind the Roo Voice desktop app.

Current file: v2 (checkpoint-2000) — TEST release. A v3 train with expanded lexical coverage is in progress and will replace this as the shipping model.

  • Reference-free: text in, one fixed voice out. Greedy decoding (temp 0).
  • Phonemization: espeak-ng en-gb, phonemizer settings must match training.
  • Decode speech codes with NeuCodec (or the int8 ONNX decoder for on-device use).
  • Speech-token mapping when serving via llama.cpp /completion + return_tokens: code = token_id − 151671; stop at 151670 (<|SPEECH_GENERATION_END|>).
FileQuantSizeNote
roo-neutts-q4_k_m.ggufQ4_K_M441 MBapp default; ear-checked
roo-neutts-q8_0.ggufQ8_0568 MBheadroom option; ear-checked

Q4_0 is deliberately absent — it degraded under greedy decoding in testing.

Sources

1 source
Hugging FacePrimaryVersion 4b22111ce98b3821767e699e03c03d83b7c1b754
huggingface.co/abliter8-ai/Roo-Voice-NeuTTS-GGUF/tree/4b22111ce98b3821767e699e03c03d83b7c1b754

Verified Oct 19

Download files

Model artifacts

4 artifacts

roo-neutts-q4_k_m.gguf

gguf · 441 MB · SHA-256 c6f55fc54b11…bb75 · Hugging Face

Download file

roo-neutts-q8_0.gguf

gguf · 568 MB · SHA-256 4ba11ea8093f…3a76 · Hugging Face

Download file

v2/roo-neutts-v2-q4_k_m.gguf

gguf · 441 MB · SHA-256 67bb81a982b4…9144 · Hugging Face

Download file

v2/roo-neutts-v2-q8_0.gguf

gguf · 568 MB · SHA-256 332d97aab0da…c54b · Hugging Face

Download file

Source excerpts

3 excerpts
ReadmeAug 3huggingface.co/abliter8-ai/Roo-Voice-NeuTTS-GGUF

--- license: apache-2.0 base_model: neuphonic/neutts-air language: [en] tags: [text-to-speech, tts, gguf, llama.cpp, single-voice] --- # Roo-Voice-NeuTTS-GGUF Single-voice identity-locked fine-tune of [NeuTTS-Air](https://huggingface.co/neuphonic/neutts-air) (Qwen2.5-0.5B backbone) — the custom voice model behind the [Roo Voice desktop app](https://github.com/abliter8-ai/roo-voice-tts). **Current file: v2 (checkpoint-2000) — TEST release.** A v3 train with expanded lexical coverage is in progress and will replace this as the shipping model. - Reference-free: text in, one fixed voice out. Greedy decoding (temp 0). - Phonemization: espeak-ng **en-gb**, phonemizer settings must match training. - Decode speech codes with [NeuCodec](https://huggingface.co/neuphonic/neucodec) (or the int8 ONNX decoder for on-device use). - Speech-token mapping when serving via llama.cpp `/completion` + `return_tokens`: code = token_id − 151671; stop at 151670 (`<|SPEECH_GENERATION_END|>`). | File | Quant | Size | Note | |---|---|---|---| | roo-neutts-q4_k_m.gguf | Q4_K_M | 441 MB | app default; ear-checked | | roo-neutts-q8_0.gguf | Q8_0 | 568 MB | headroom option; ear-checked | Q4_0 is deliberately absent — it degraded under greedy decoding in testing.

NoteOct 19huggingface.co/abliter8-ai/Roo-Voice-NeuTTS-GGUF

Source context: 202 downloads · 0 likes · Pipeline text-to-speech · Repo abliter8-ai/Roo-Voice-NeuTTS-GGUF

NoteAug 3huggingface.co/abliter8-ai/Roo-Voice-NeuTTS-GGUF

Source context: 595 downloads · 0 likes · Pipeline text-to-speech · Repo abliter8-ai/Roo-Voice-NeuTTS-GGUF