Fast and multilingual

Use Piper for fast multilingual text to speech

Piper is the practical choice when you need languages beyond English or want a separate, relatively small download for each voice. The current AIVoices studio includes 48 Piper voices across 24 language labels and runs them through WebAssembly on the CPU.

Your text is processed locally by the selected model.0 / 5000

Choose Piper for language coverage and modest downloads

Piper works well for reading support, long documents, utility speech, and multilingual projects where clarity and CPU compatibility matter more than expressive delivery. Each voice downloads separately, so you only store the voices you actually use.

Test the exact regional voice you plan to keep

  1. 1Open the relevant language page and choose the regional voice that matches the audience.
  2. 2Generate a representative passage with names, dates, and any words borrowed from another language.
  3. 3Keep the same voice across the project and export corrected sections as clearly labelled WAV files.

Specifications

DeveloperRhasspy / Michael Hansen
Parameters20 million
Download size27-115 MB per voice (fp16 per-voice)
Output22.1 kHz mono
Browser voices48
LanguagesEnglish (US), English (UK), Spanish, French, German, Italian, Portuguese (BR), Russian, Chinese, Dutch, Polish, Turkish, Ukrainian, Czech, Danish, Norwegian, Swedish, Finnish, Greek, Hungarian, Romanian, Vietnamese, Arabic, Catalan
Voice cloningNo
StreamingNo
LicenseMIT (core) — The Piper code is MIT licensed. Individual voice files may use different dataset licenses, so check the selected voice before commercial use.
SpeedDesigned for fast CPU generation without requiring a dedicated GPU.
RuntimeWebAssembly (CPU)

Specifications checked 2026-07-21 against:

Explore related downloads

The browser studio and downloadable catalog are separate. Use the catalog to find independently published model files and check each source before downloading.

Good fit when

  • +Wide language and voice coverage
  • +Small download for each selected voice
  • +Fast generation on ordinary CPUs

Before you choose it

  • Voices can sound more synthetic than larger models
  • Each voice is a separate model and does not support voice references

Voices (48)

HFC Female · en-USRyan · en-USAmy · en-USLessac · en-USAlan · en-GBCori · en-GBDaveFX · es-ESSharvard · es-ESALD (México) · es-MXClaude (México) · es-MXSiwis · fr-FRUPMC · fr-FRTom · fr-FRThorsten · de-DEThorsten (emotional) · de-DEMLS · de-DERiccardo · it-ITFaber (Brasil) · pt-BREdresson (Brasil) · pt-BRTugão (Portugal) · pt-PTИрина · ru-RUДмитрий · ru-RUДенис · ru-RUРуслан · ru-RU华严 · zh-CNكريم · ar-JOGosia · pl-PLDarkman · pl-PLMC Speech · pl-PLMLS · nl-NLNathalie (Vlaams) · nl-BERDH (Vlaams) · nl-BEDFKI · tr-TRFahrettin · tr-TRFettah · tr-TRUkrainian TTS · uk-UAJirka · cs-CZNST · sv-SEVAIS 1000 · vi-VNMihai · ro-ROAnna · hu-HUImre · hu-HUBerta · hu-HUHarri · fi-FITalesyntese · da-DKTalesyntese · no-NOامیر · fa-IRژیرو · fa-IR

See it alongside another model

Questions about Piper

Does AIVoices charge to use Piper?

AIVoices does not charge generation credits for this browser model. The listed model license is MIT (core). The Piper code is MIT licensed. Individual voice files may use different dataset licenses, so check the selected voice before commercial use.

What hardware does Piper need?

A current desktop browser and CPU are enough for this model. The 27-115 MB per voice model is stored in your browser after the first download.

Is my text private when using Piper here?

The supported model executes inside your browser tab. AIVoices does not send the text to a server for speech generation.

Can I use Piper output commercially?

Commercial use depends on the model, selected voice, and source-data terms. This page lists MIT (core) for the model and notes that The Piper code is MIT licensed. Individual voice files may use different dataset licenses, so check the selected voice before commercial use.