Skip to content
Littlewolf Labs News

Paywall Dodgers

AI voices

Speak or clone a voice on your own computer instead of paying for a hosted studio. Cloning and plain readout are listed apart.

Badge key
  • Fully free / open source. A standard open-source license.
  • Free tier. The product has a free plan.
  • Free app, paid API. The app is free to install. Making something still costs money.
  • Unofficial: use at your own risk. Someone else built this. It can break. Back up your files.
  • Early. Not a finished replacement.
  • Not yet tested by us. Listed from the project page. Not a hands-on test.
  • Tested by us. We ran it on our own machine. The card has the result.

Voice cloning

Copy a voice from a short sample on a computer you own. You need permission from the person whose voice you copy.

What it costs

  • ElevenLabs

    • Free $0/month
    • Starter $6/month, or $5/month if you pay $60 up front for the year
    • Creator $22/month, or $18.33/month if you pay $220 up front for the year
    • Pro $99/month, or $82.50/month if you pay $990 up front for the year
    • Scale $299/month, or $249.17/month if you pay $2,990 up front for the year
    • Business $990/month, or $825/month if you pay $9,900 up front for the year

    List prices checked , Pacific time. Official pricing page.

Free alternatives that best fit

Best overall free pick: ChatterboxClosest local clone: GPT-SoVITS

Chatterbox

  • Fully free / open source
  • Tested by us

Chatterbox is Resemble AI's own open voice model, for people who want local speech from the company that trained it. Turbo and Nano are smaller versions of the same project. You run it on your computer, and you still supply the machine.

Tested on CPU only. The original model ran about 4 times slower than real time, and Nano ran faster than real time. The published package needed an older setuptools version to install.

Replaces
ElevenLabs
Price
Free
License
MITPublished by Resemble AI. The same repository includes the original model plus the smaller Turbo and Nano versions.
Hardware
CPU or a graphics card. Turbo and Nano are the smaller versions.
Maintained
Updated July 21, 2026.

Repository

GPT-SoVITS

  • Fully free / open source
  • Not yet tested by us

GPT-SoVITS copies a voice from a short sample on a computer you own. It fits people who want a local clone instead of a hosted studio. Speaking can run on a CPU. Teaching it a new voice needs a graphics card, and you need consent before you clone anyone.

Replaces
ElevenLabs
Price
Free
License
MIT
Hardware
Runs on CPU for inference. Fine-tuning needs a GPU with 8 to 14 GB.
Maintained
Updated October 8, 2026.

Repository

Speech without cloning

Read text aloud in a stock voice. These speak in their own voices.

What it costs

  • ElevenLabs

    • Free $0/month
    • Starter $6/month, or $5/month if you pay $60 up front for the year
    • Creator $22/month, or $18.33/month if you pay $220 up front for the year
    • Pro $99/month, or $82.50/month if you pay $990 up front for the year
    • Scale $299/month, or $249.17/month if you pay $2,990 up front for the year
    • Business $990/month, or $825/month if you pay $9,900 up front for the year

    List prices checked , Pacific time. Official pricing page.

Free alternatives that best fit

Best maintained: Coqui TTS (Idiap fork)Lightest CPU: Piper

Piper

  • Fully free / open source
  • Tested by us

Piper is a small local voice for reading text on a CPU, the lightest option here. The Open Home Foundation keeps the current repository, and its README says they are looking for more maintainers. It speaks in stock voices. It does not clone a person.

Tested on CPU. Six seconds of speech in under a quarter second.

Replaces
ElevenLabs
Price
Free
License
GPL 3.0The current project is OHF-Voice/piper1-gpl. The older rhasspy/piper repository is archived and was MIT.
Hardware
A CPU. No graphics card.
Maintained
Updated October 16, 2026.The README says the Open Home Foundation is looking for maintainers.

Repository

Coqui TTS (Idiap fork)

  • Fully free / open source
  • Not yet tested by us

Coqui TTS is Idiap's local text-to-speech library, for people who want a maintained stock voice. You run it on your own computer. The XTTS-v2 voice weights are a separate, non-commercial license.

Replaces
ElevenLabs
Price
Free
License
Mozilla Public License 2.0The library is MPL 2.0. The XTTS-v2 weights use the Coqui Public Model License, which is non-commercial.
Hardware
Your own computer
Maintained
Updated October 16, 2026.

Repository

Kokoro

  • Fully free / open source
  • Tested by us

Kokoro is a small text-to-speech model, about 82 million parameters, for plain readout on a modest computer. It fits people who want something light. It has not moved since August 2025.

Tested Oct 16, 2026 on an 8-core CPU with no graphics card. It spoke a sentence in about 1 second and a 45-second passage in about 7 seconds, and Whisper read it back nearly word for word.

Replaces
ElevenLabs
Price
Free
License
Apache 2.0
Hardware
A small model on an ordinary computer
Maintained
Last update August 6, 2025. Quiet for the six months before October 16, 2026.

Repository