Piper TTS setup on Jetson Orin Nano with JetPack 7.2

This tutorial sets up Piper TTS on an NVIDIA Jetson Orin Nano running JetPack 7.2. The speaker presents it as an offline speech output stage for a future voice assistant connected to a local LLM. The lesson demonstrates text-to-speech; speech recognition and model integration remain future steps.

Setup starts with a system update and an audio development dependency, followed by downloading and unpacking Piper. Each voice needs an ONNX model and its matching JSON file. The Python script uses explicit paths to the Piper executable and voice model, so viewers must adjust the username and update the model path when choosing another voice.

The audio configuration gets particular attention. In the demonstrated pipeline, the speaker explicitly selects an output device rather than relying on the desktop sound setting. He lists audio devices with aplay and uses card zero, device zero for his USB speakerphone. Playback uses raw audio at 22,050 Hz in S16LE format. Python sends UTF-8 text through a subprocess pipe to Piper, whose output feeds aplay.

The demonstration produces spoken messages, though the speaker dislikes the selected voice. He suggests trying other voices and turning the script into an echo bot with keyboard input. The Jetson needs an external speaker, and Bluetooth playback is untested in this setup. A microphone is relevant to the planned speech recognition lesson.

The description discloses Amazon affiliate links for the hardware.