OpenVoiceOS/phoonnx-outetts

phoonnx OuteTTS 1.0 mirror

ONNX artifacts for the OuteTTS 1.0 family, laid out for the outetts engine in phoonnx. All models are redistributed unchanged from the official OuteAI and IBM Research repositories. This mirror adds no weights of its own; it only guarantees a stable file layout and stable URLs for phoonnx voice entries.

Contents

Path	Source	License
0.6B/model.onnx (+ model.onnx_data), 0.6B/tokenizer.json	OuteAI/OuteTTS-1.0-0.6B-ONNX	Apache-2.0
1B/model.onnx (+ model.onnx_data)	re-exported by phoonnx from OuteAI/Llama-OuteTTS-1.0-1B - see below	CC-BY-NC-SA-4.0 (non-commercial)
1B/tokenizer.json	OuteAI/Llama-OuteTTS-1.0-1B-ONNX	CC-BY-NC-SA-4.0 (non-commercial)
dac/decoder_model.onnx, dac/encoder_model.onnx, dac/config.json	OuteAI/DAC-speech-v1.0-ONNX (weights: ibm-research/DAC.speech.v1.0)	CDLA-Permissive-2.0
speakers/en-female-1-neutral.json	the outetts Python package (version/v3/default_speakers)	Apache-2.0

The two model sizes carry different licenses. The 0.6B is Apache-2.0 and covers 14 languages; phoonnx resolves those 14 to it. The 1B is CC-BY-NC-SA-4.0 - not licensed for commercial use - and is here because it is the only member of the family that covers Arabic, Belarusian, Bengali, Lithuanian, Persian, Portuguese, Swahili, Tamil and Ukrainian. Check the 1B license before you deploy it.

Model authorship and all research credit belong to OuteAI. The audio codec is DAC.speech.v1.0 by IBM Research.

Languages

• 0.6B (Apache-2.0), 14 trained languages: English, Chinese, Dutch, French, Georgian, German, Hungarian, Italian, Japanese, Korean, Latvian, Polish, Russian, Spanish.
• 1B (CC-BY-NC-SA-4.0), 23 trained languages: the 14 above plus Arabic, Belarusian, Bengali, Lithuanian, Persian, Portuguese, Swahili, Tamil, Ukrainian.

Numeric parity

Every export in this mirror was checked against the upstream torch weights on a fixed 1843-token prompt (greedy, 64 decode steps, CPU float32).

Export	max abs logit diff	greedy agreement	shipped
0.6B model.onnx (fp32)	3.8e-05	64/64	yes
0.6B model_fp16.onnx	0.036	63/64	no
0.6B model_quantized / uint8	5.7	62/64	no
0.6B model_bnb4 / int8 / q4 / q4f16	3.9 - 5.9	32-34/64	no
1B model.onnx - OuteAI's (fp32)	12	61/64	no, see below
1B model.onnx - this repo's re-export (fp32)	1.78e-05	64/64	yes

The 0.6B float32 export reproduces its torch weights exactly. Every quantized 0.6B export changes greedy decoding, so none is a safe default.

The official 1B ONNX export is broken; this one is not

OuteAI/Llama-OuteTTS-1.0-1B-ONNX does not reproduce its own torch weights, even in float32. On the prompt above its last-position logits are off by 12, the logit correlation is 0.48, prefill argmax agrees on only 1580 of 1843 positions, and greedy decoding diverges from OuteAI/Llama-OuteTTS-1.0-1B. The error is already present on a 32-token prompt (0.08 max diff) and never closes (correlation 0.96-0.98 at 512-2048 tokens), so it is a property of the export, not of long-context accumulation. The 0.6B export on the same prompt differs by 3.8e-05. Every quantized 1B export inherits the defect and adds to it.

The 1B/model.onnx in this repository is therefore not OuteAI's file. It was re-exported from the torch checkpoint with scripts/conversion/outetts/export_outetts_onnx.py and verified the same way: max abs logit diff 1.78e-05, mean 3.74e-06, logit correlation 1.00000000, greedy 64/64, and 64/64 again under the windowed repetition penalty OuteTTS actually samples with. The same script reproduces the 0.6B to 1.6e-05, which is how it is checked against a graph already known to be good.

One difference to know about: the re-exported graph returns logits shaped [1, 1, V] - the final row only - where OuteAI's exports return every position. Any consumer that reads the last row, as transformers.js and phoonnx both do, works with either.

Use it from phoonnx

from phoonnx.model_manager import TTSModelManager

manager = TTSModelManager()
manager.merge_default_voices()
voice = manager.voices["outetts/0.6B/en"].load()
for chunk in voice.synthesize("Hello from OuteTTS running in phoonnx."):
    ...  # chunk.audio_float_array, 24 kHz mono

Voice ids run outetts/0.6B/<lang> for the 14 languages the 0.6B was trained on and outetts/1B/<lang> for the nine only the 1B covers (ar, be, bn, fa, lt, pt, sw, ta, uk). Every one of them resolves to this repository. The engine drives the LM with a KV cache, then decodes the two DAC codebooks to a waveform.

How it works

The LM emits interleaved <|c1_N|><|c2_N|> tokens. dac/decoder_model.onnx takes audio_codes[1, 2, T] and returns audio_values[1, 1, T * 512] at 24 kHz. A speaker profile (speakers/*.json) is an in-context audio prompt: a transcript whose words each carry a duration, three prosody buckets and their DAC codes.

Ethical use

OuteAIs guidance applies unchanged: do not clone a voice without the speakers explicit permission, and do not use this model to impersonate or to produce deceptive content.