| name | sherpa-onnx-tts |
| description | Local offline text-to-speech using Sherpa ONNX with Piper voices |
| emoji | 🗣 |
| version | 1.0.0 |
| requires | {"bins":["sherpa-onnx-offline-tts"]} |
| invocation | {"userInvocable":true} |
Sherpa ONNX TTS (Local Offline Text-to-Speech)
Convert text to speech audio entirely offline using the Sherpa ONNX runtime with Piper voice models. No cloud API or internet connection required.
Usage
- "Say 'Hello, welcome to the meeting' and save as audio"
- "Convert this paragraph to speech"
- "Generate an audio file of these instructions using a female voice"
- "Read this text aloud at 1.2x speed"
How it works
Runs the sherpa-onnx-offline-tts binary, which loads a Piper ONNX voice model and synthesizes speech from input text. Output is a WAV audio file.
Key flags
--vits-model=<path> -- path to the ONNX voice model file
--vits-tokens=<path> -- path to the tokens file for the model
--vits-data-dir=<path> -- path to the espeak-ng-data directory (for Piper models)
--output-filename=<path> -- output WAV file path
--speed=<float> -- speech speed multiplier (default 1.0; higher is faster)
--sid=<int> -- speaker ID for multi-speaker models
Example command
sherpa-onnx-offline-tts \
--vits-model=./models/en_US-amy-medium.onnx \
--vits-tokens=./models/tokens.txt \
--vits-data-dir=./models/espeak-ng-data \
--output-filename=output.wav \
--speed=1.0 \
"Hello, this is a test of local text to speech."
Available voices
Piper provides dozens of voices across many languages. Popular English voices include:
en_US-amy-medium (female, US English)
en_US-joe-medium (male, US English)
en_GB-alan-medium (male, British English)
Voice models can be downloaded from https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
Setup
- Install sherpa-onnx: download pre-built binaries from https://github.com/k2-fsa/sherpa-onnx/releases or build from source.
- Download at least one Piper voice model (ONNX file + tokens file + espeak-ng-data).
- Ensure
sherpa-onnx-offline-tts is on your PATH.