mms-tts-orm-onnx

Oromo (Afaan Oromoo) text-to-speech, converted for sherpa-onnx from Meta's MMS-TTS checkpoints.

Model properties

Read from the metadata embedded in model.onnx:

Property	Value
model_type	vits
language	orm
frontend	characters
n_speakers	0 (single speaker)
sample_rate	16000 Hz

The frontend is character-based, so no lexicon or espeak-ng data is needed — tokens.txt is the full symbol table.

Files

File	Description
model.onnx	VITS acoustic model + vocoder (~109 MiB)
tokens.txt	Character-to-id symbol table
sample.wav	Reference output, 16 kHz mono

Usage

sherpa-onnx-offline-tts \
  --vits-model=./model.onnx \
  --vits-tokens=./tokens.txt \
  --output-filename=./out.wav \
  "your text here"

Python:

import sherpa_onnx, soundfile as sf

tts = sherpa_onnx.OfflineTts(
    sherpa_onnx.OfflineTtsConfig(
        model=sherpa_onnx.OfflineTtsModelConfig(
            vits=sherpa_onnx.OfflineTtsVitsModelConfig(
                model="model.onnx", tokens="tokens.txt"
            ),
        ),
    )
)
audio = tts.generate("your text here")
sf.write("out.wav", audio.samples, audio.sample_rate)

License

Inherited from the upstream MMS checkpoints: CC-BY-NC 4.0 — non-commercial use only.

Upstream: https://huggingface.co/facebook/mms-tts/tree/main