# Piper

Speech you host. Piper is GPL-3.0 — ElevenLabs without the character meter.

**Replaces:** ElevenLabs

**Tier:** Easy

## Problem

ElevenLabs meters characters for hosted voices. You wanted text-to-speech on a box you run. Piper (Open Home Foundation, GPL-3.0) is a local neural TTS engine with an HTTP API. Each voice file has its own license — read the model card. This is not ElevenLabs voice cloning or every studio voice. Coqui TTS is archived; do not start there.

## Stack

- [Piper](https://github.com/OHF-Voice/piper1-gpl) — GPL-3.0 local neural text-to-speech

## Estimated SaaS cost saved

| You stop paying | Typical / year | Self-host instead |
|---|---:|---|
| ElevenLabs paid (small team) | $1,200 | Piper + voices on a volume |

**About $1,200 / year** on a small team, before you count lock-in.

## Deploy on Railway

Placeholder: Piper HTTP API plus a volume for voice models. CPU is enough for many voices. No one-click template in this repo yet — Start on Railway.

[Start on Railway](https://c17.io/go/railway/piper-tts?from=%2Fstacks%2Fpiper-tts.md)

[Open WebUI if this is chat, not speech](https://c17.io/stacks/open-webui-litellm)

## AGENT_PROMPT

```
Replace ElevenLabs with Piper (https://github.com/OHF-Voice/piper1-gpl).

- Run the HTTP API. Persist voices on a volume. Synthesize one sentence you actually ship.
- Check each voice model's license. Do not claim cloning or ElevenLabs studio parity.
- Write AGENT_PROMPT.md.
```

## FAQ

### Can it clone a voice?

Not the ElevenLabs product. Piper speaks trained voices you download. Read each voice license.

### Why not Coqui?

Coqui TTS is archived. Piper is the maintained local engine.

Source: https://c17.io/stacks/piper-tts