abliter8-ai/Roo-Voice_MOSS_TTS_LT_mlx8

Roo-Voice · MOSS-TTS-Local-Transformer · MLX 8-bit

Roo's voice — that signature baritone with the estuary accent that stands the hair up on the back of your neck — as an 8-bit MLX model that runs on Apple Silicon. Load it, and Roo can whisper to you all day long.

An 8-bit MLX quantization of a full-model supervised fine-tune of MOSS-TTS-Local-Transformer, trained on Roo's own recordings and specialised on his single voice.

What actually makes the voice — read this

This is a reference-conditioned model, and both halves matter:

• The fine-tune is what makes it Roo. The base model has never heard this speaker — a base model plus any reference clip will not give you Roo's baritone or his estuary accent. That voice lives in the weights, put there by the supervised fine-tune on his recordings.
• The reference completes the delivery. reference.wav (bundled) conditions the fine-tuned model at inference and is required to produce the voice.

So the product is this fine-tune + reference.wav, together — neither the base model with a reference, nor this checkpoint without one, reproduces Roo. It is not a text-only model, and it is not a generic voice-cloner: swap in a different reference and you are not getting Roo, because the accent and timbre are the fine-tune's, not the clip's.

What it is

Voice	Single speaker — Roo (baritone, estuary accent), 24 kHz mono
Format	MLX, 8-bit affine weights, group size 64, BF16 retained (W8Abf16)
Runtime	mlx-audio on Apple Silicon
Weights	model.safetensors ≈ 3.6 GB
Source	Full-model SFT (not LoRA/adapter); 556 tensors
Base	OpenMOSS-Team/MOSS-TTS-Local-Transformer @ 12aa734e4f11a7b3fdf4eb0ad2aa2029675ffc2e
Audio codec	OpenMOSS-Team/MOSS-Audio-Tokenizer @ 3cd226ba2947efa357ef453bcad111b6eafba782 (fetched by mlx-audio)

Usage

import mlx.core as mx
from mlx_audio.tts.utils import load_model

model = load_model("./")  # this repo

mx.random.seed(42)
result = None
for r in model.generate(
    text="After the last dance class, I parked the car beside the garden wall.",
    ref_audio="reference.wav",          # REQUIRED — conditions the fine-tuned Roo voice
    mode="generation",
    max_tokens=4096,
    n_vq_for_inference=32,
    text_temperature=1.0, text_top_p=0.95, text_top_k=50, text_repetition_penalty=1.0,
    audio_temperature=1.0, audio_top_p=0.95, audio_top_k=50, audio_repetition_penalty=1.1,
):
    result = r
# result.audio -> 24 kHz mono float

Decoding contract for this voice: seed 42, temperature 1.0, top-k 50, top-p 0.95, repetition penalty 1.1, 32 RVQ codebooks.

Limitations

• Reference-conditioned — the bundled reference.wav must ride along; there is no text-only path.
• Single voice by design (this is Roo, not a multi-speaker system).
• 8-bit quantization: a small quality delta vs the FP32/BF16 source is possible.

Provenance & license

Quantized/exported form of an accepted single-speaker MOSS-TTS Local supervised fine-tune. The base model and audio codec are Apache-2.0 (OpenMOSS); weights derived from them are redistributed here under the same license.