Piper TTS
A fast, CPU-only neural TTS engine from Rhasspy that runs entirely offline, even on a Raspberry Pi.
Piper was built by the Rhasspy project to bring neural text-to-speech to edge devices that have no GPU. It pairs a VITS-style synthesizer with the Larynx pipeline and runs as a compact ~15M-parameter ONNX model directly on CPU, so synthesis happens locally with no cloud round-trip. Its strength is breadth and footprint: over 100 voices spanning 35+ languages at real-time speed on hardware as modest as a Pi 4. That makes it the practical default for home automation, accessibility readers, and offline kiosks where latency, privacy, and zero hosting cost matter more than studio polish.
At a glance
- Developer
- Rhasspy
- License
- MIT
- Tier
- free
- Speed
- fast
- Voice cloning
- No
- Languages
- English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Chinese, Arabic, Czech, Danish, Finnish, Greek, Hungarian, Icelandic, Georgian, Kazakh, Nepali, Norwegian, Romanian, Slovak, Serbian, Swedish, Swahili, Turkish, Ukrainian, Vietnamese, CA, CY, FA, Latvian, Slovenian, LB, Basque, Indonesian, Kurdish, Malayalam, Albanian, Telugu, Urdu
- Max characters
- 2000
Piper AI Voices
Best for
Quick previews, accessibility, and embedded applications
Piper TTS — FAQ
No. Piper is CPU-only by design and uses 0 VRAM — it runs offline on embedded hardware like a Raspberry Pi 4 in real time, which is exactly what it was built for.
Yes. Piper is released under the MIT license, so it is free to use in commercial and personal projects.
Piper offers 100+ voices across 35+ languages, including English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Chinese, Arabic, Turkish, Ukrainian and many more. It supports SSML and handles up to 2,000 characters per request.