saasisdead.si
R.I.P.

Open-source alternatives to ElevenLabs

6 projects replace ElevenLabs, the AI voice generator. 5 shipped code in the last 30 days.

ElevenLabs Starter$6/molist price

ElevenLabs sells credits by the character, so audiobooks, dubbing and voice agents get expensive. Voices and voice clones live on its platform under its terms. Open speech models have improved quickly since 2024, and many creators now run them on their own GPU.

Chatterbox, from Resemble AI, clones a voice from a short clip and lets you turn emotion up or down. Kokoro is a small model that runs quickly even on a CPU, good for narration in bulk. Fish Speech and F5-TTS clone voices from a few seconds of reference audio. Most of these ship as model weights with a script or simple demo, not as a finished studio.

6 open-source alternatives

Most-starred first. Stars, growth and activity are live from GitHub.

GPT-SoVITSMost adopted

1 min voice data can also be used to train a good TTS model! (few shot voice cloning)

Stars
63.3k
Growth
–
Launched
Jan 2024
MITupdated 1d ago
Fish Speech#2 by stars

SOTA Open Source TTS

Stars
33.3k
Growth
–
Launched
Oct 2023
Custom licenseupdated 6d ago
ChatterboxNewest contender

SoTA open-source TTS

Stars
27.1k
Growth
–
Launched
Apr 2025
MITupdated 3mo ago
IndexTTS#4 by stars

An Industrial-Level Controllable and Efficient Zero-Shot Text-To-Speech System

Stars
24.6k
Growth
–
Launched
Feb 2025
Custom licenseupdated 21d ago
F5-TTS#5 by stars

Official code for "F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching"

Stars
15.4k
Growth
–
Launched
Oct 2024
MITupdated 6d ago
Piper#6 by stars

Fast and local neural text-to-speech engine

Stars
5.9k
Growth
–
Launched
Mar 2025
GPL-3.0updated 3d ago

What you give up by leaving ElevenLabs

The honest part. Free and open usually costs something else.

  • Not a finished app Most projects ship weights and a demo, so you build the studio, API and storage yourself.
  • Language coverage Open models are strongest in English and Chinese and less reliable in many other languages.
  • GPU for cloning Most cloning and expressive models need a decent GPU to run faster than real time.
  • License limits Some model weights forbid commercial use, so read each license before you publish audio.

All AI & Agents

Esc

Type to search every entry. ↑↓ to move, Enter to open.