mannix/ outetts

204 Downloads Updated 1 month ago

OuteTTS 0.3 500M · WavTokenizer vocoder

ollama run mannix/outetts:0.3

curl http://localhost:11434/api/chat \
  -d '{
    "model": "mannix/outetts:0.3",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='mannix/outetts:0.3',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'mannix/outetts:0.3',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size / Usage

Context

Input

outetts:0.3

667MB · - context window · Text · 1 month ago

outetts:0.3

667MB

-

Text

Readme

OuteTTS 0.3 500M — text to speech, for xOllama

OuteTTS 0.3 500M (Q8_0 GGUF) · WavTokenizer vocoder · OpenAI /v1/audio/speech

Runs on xOllama, not on stock ollama. Stock ollama pulls this model but cannot show or run it: it carries only media engines. xOllama serves it through an opencoti engine with media support (dev build 2610022033001/b105 or later, set with XOLLAMA_ENGINE_PATH). The engine pinned by the current xOllama release has no media engines and refuses the model at load, naming what it lacks.

Speech from a 500M language model and the WavTokenizer vocoder: the pair opencoti’s first media engine was tested with. For more voices or less memory, see Kokoro, Supertonic and KittenTTS.

Use

xollama pull mannix/outetts:0.3
curl http://localhost:22434/v1/audio/speech \
  -d '{"model": "mannix/outetts:0.3", "input": "The first ship came in at dawn."}' -o speech.mp3

The audio comes back as the engine encodes it: mp3 by default on engines with the codec (b103 and later), wav before. xollama media voices mannix/outetts:0.3 lists the voices.

Measured

	RTX 3090 (b97)
one sentence	about 4 s; transcribed back exactly by Whisper

Tags

Tag	Weights	Size
`0.3`	`OuteTTS-0.3-500M-Q8_0.gguf` + `WavTokenizer-Large-75-F16.gguf`	667 MB

Source

OuteAI/OuteTTS-0.3-500M-GGUF · vocoder: ggml-org/WavTokenizer · License: CC-BY-SA-4.0. Docs: xOllama media.

# OuteTTS 0.3 500M — text to speech, for xOllama

**OuteTTS 0.3 500M (Q8_0 GGUF) · WavTokenizer vocoder · OpenAI `/v1/audio/speech`**

> **Runs on [xOllama](https://github.com/mann1x/xollama), not on stock ollama.** Stock ollama pulls this model but cannot show or run it: it carries only media engines. xOllama serves it through an [opencoti](https://huggingface.co/ManniX-ITA/opencoti-llamafile) engine with media support (dev build `2610022033001`/b105 or later, set with `XOLLAMA_ENGINE_PATH`). The engine pinned by the current xOllama release has no media engines and refuses the model at load, naming what it lacks.

Speech from a 500M language model and the WavTokenizer vocoder: the pair opencoti's first media engine was tested with. For more voices or less memory, see [Kokoro](https://ollama.com/mannix/kokoro), [Supertonic](https://ollama.com/mannix/supertonic) and [KittenTTS](https://ollama.com/mannix/kittentts).

## Use

```shell
xollama pull mannix/outetts:0.3
curl http://localhost:22434/v1/audio/speech \
  -d '{"model": "mannix/outetts:0.3", "input": "The first ship came in at dawn."}' -o speech.mp3
```

The audio comes back as the engine encodes it: mp3 by default on engines with the codec (b103 and later), wav before. `xollama media voices mannix/outetts:0.3` lists the voices.

## Measured

| | RTX 3090 (b97) |
|---|---|
| one sentence | about 4 s; transcribed back exactly by Whisper |

## Tags

| Tag | Weights | Size |
|---|---|---|
| `0.3` | `OuteTTS-0.3-500M-Q8_0.gguf` + `WavTokenizer-Large-75-F16.gguf` | 667 MB |

## Source

[OuteAI/OuteTTS-0.3-500M-GGUF](https://huggingface.co/OuteAI/OuteTTS-0.3-500M-GGUF) · vocoder: [ggml-org/WavTokenizer](https://huggingface.co/ggml-org/WavTokenizer) · License: CC-BY-SA-4.0.
Docs: [xOllama media](https://github.com/mann1x/xollama/blob/dev/docs/xollama/media.mdx).

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)