otter-vox

otter-vox turns text into speech on your machine with a native Audio8 ggml runtime. Vulkan when it is available. CPU otherwise. No Python, ONNX, or CUDA.

otter-vox ships in otter-shell-extras, not the base otter-shell metapackage. Extras also pulls otter-vox-model, the Audio8 GGUF and tokenizer under /usr/share/otter-shell/models/vox. Install with pikman install otter-shell-extras.

Usage

otter-vox --play 'Welcome to Otter Shell.'
printf 'Welcome to Otter Shell.' | otter-vox --no-play > speech.wav
otter-vox --no-play -o speech.wav 'Welcome to Otter Shell.'
otter-vox --list-voices
otter-vox --list-gpus
otter-vox --check-model

Keep the model warm for repeated speech:

otter-vox --serve

Send one utterance per line on stdin while --serve is running.

Voices

Default voice is fer (pre-encoded zero-shot clone). Fallback is audio8-en-calm from Audio8’s licensed English reference. Select it with --voice audio8-en-calm. Audio8 supports Cantonese, Chinese, Dutch, English, French, German, Italian, Japanese, Korean, Polish, and Spanish text. Voice identity transfers across those languages.

Backend and GPU

otter-vox --cpu --play 'CPU only.'
otter-vox --backend vulkan --gpu gpu0 --play 'First discrete GPU.'

Config / flag	Meaning
`backend = auto`	Prefer Vulkan on NVIDIA, AMD, and Intel; else CPU
`backend = vulkan` / `cpu`	Force a backend (`--cpu` forces CPU)
`gpu = auto`	Prefer a dedicated GPU, else the main available one
`gpu = gpu0` … `gpu15`	Pin a device from `otter-vox --list-gpus`
`cpu_threads = 8`	CPU thread count (capped to available logical CPUs)

Vulkan playback uses a non-blocking PipeWire queue with a 2-second prebuffer. CPU output buffers before playback so you do not get gaps.

Configuration

~/.config/otter-shell/otter-vox.conf (also editable under Voice in otter-settings):

model_dir = ""
output_path = "-"
playback = true
voice = "fer"
backend = auto
gpu = auto
cpu_threads = 8

Empty model_dir uses the packaged GGUF under /usr/share/otter-shell/models/vox. The shipped model is a Q5_0/FP16 conversion of Audio8 TTS Preview 0.6B (~643 MiB).

Get Started

Desktop Shell

Core Apps

Productivity

Utilities

Gaming

Configuration

Usage

Voices

Backend and GPU

Configuration

​Usage

​Voices

​Backend and GPU

​Configuration

Usage

Voices

Backend and GPU

Configuration