Implement a voice translation pipeline on Apple silicon
Introduction
Set up the voice translation project
Transcribe microphone audio with Sherpa-ONNX
Translate text with Gemma 4 E2B
Synthesize translated speech with Supertonic 3
Build the voice translator with Gradio
Validate the voice translation pipeline
Next Steps
Implement a voice translation pipeline on Apple silicon
Who is this for?
This Learning Path is for Python developers who want to implement a voice translation pipeline locally on an Apple silicon Mac. You'll translate English speech into Spanish. You can select other supported target languages, but recognizing speech in another source language requires a different speech recognition model and configuration.
What will you learn?
After completing this Learning Path, you'll be able to perform the following tasks:
- Prepare LiteRT-LM and a Gemma 4 model for local translation on an Apple silicon Mac.
- Capture microphone input and transcribe English speech locally with Sherpa-ONNX.
- Prompt Gemma 4 to translate the English transcript into Spanish.
- Synthesize and play the Spanish translation with Supertonic 3 and Sherpa-ONNX.
Prerequisites
Before starting, you'll need the following prerequisites:
- A Mac with Apple silicon running macOS 12 or later, with a microphone, speakers, at least 8 GB of memory, and 15 GB of free storage
- Python 3.10 through 3.13
- Internet access for dependency and model downloads
Summary
This summary was drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.
You’ll build a local English-to-Spanish voice translator on an Apple silicon Mac. First, you’ll prepare Python and download the models. You’ll then test microphone transcription with Sherpa-ONNX, text translation with Gemma 4 E2B and LiteRT-LM, and speech synthesis with Supertonic 3. Finally, you’ll connect these stages in Gradio and verify the transcript, translation, spoken result, and offline operation.