⚡ ElevenLabs Killer Just Dropped — 100% Free & Open Source

Clone Any Voice in 3 Seconds — Free Forever

Alibaba just killed the $22/month voice cloning industry.


:world_map: What You’re Walking Away With

Free, local, unlimited voice cloning that runs on a potato GPU. No account. No API keys. No “credits remaining.”


Why This Is Insane

ElevenLabs charges monthly for this exact feature.
This runs on under 2GB VRAM.
Your voice data never leaves your machine.


The Good Stuff

  • :studio_microphone: Clone any voice from 3 seconds of audio
  • :sparkles: Create voices by describing them in plain English
  • :performing_arts: Control emotion, tone, speed
  • :globe_showing_europe_africa: 10 languages out the box
  • :floppy_disk: Runs 100% offline after setup
  • :scroll: Apache 2.0 — use it for anything

Try It Now (Literally 0 Install)

:backhand_index_pointing_right: https://huggingface.co/spaces/Qwen/Qwen3-TTS

Click “Voice Clone” tab → record 3 seconds → paste what you want it to say → generate.

That’s the whole thing. You’re done. Go play.


:magic_wand: The Sleeper Feature Nobody’s Talking About

Voice Design — you describe a voice in words, it creates it from nothing.

“25 year old male, raspy, confident, slight southern accent”

And it just… makes that voice. No sample needed. Describe → generate → use.

This is genuinely black magic.


Run It Locally

3 Commands — Copy Paste This
conda create -n qwen3-tts python=3.12 -y && conda activate qwen3-tts
pip install qwen-tts
qwen-tts-demo Qwen/Qwen3-TTS-12Hz-1.7B-Base --port 8000

Open localhost:8000. Same interface as the online demo, but offline forever.

Windows error? Add --no-flash-attn to the last command.

Want voice design instead? Swap Base for VoiceDesign in the command.


Quick Fixes

If Something Breaks
Problem Fix
FlashAttention error Add --no-flash-attn
Python 3.13+ fails Use 3.12 specifically
Weird output quality Use “Ryan” voice, keep text short
Mac M-series Use this ComfyUI fork

For Tinkerers

API Server, CLI Tool, ComfyUI Nodes

Simple CLIhttps://github.com/daliusd/qtts

./qtts.py "Hello world" -s Ryan -l English

REST API + Dockerhttps://github.com/ValyrianTech/Qwen3-TTS_server

docker run --gpus all -p 7860:7860 qwen3-tts_server

ComfyUI Nodeshttps://github.com/DarioFT/ComfyUI-Qwen3-TTS


Links

:hugs: Modelshttps://huggingface.co/collections/Qwen/qwen3-tts

:video_game: Demohttps://huggingface.co/spaces/Qwen/Qwen3-TTS

:memo: Bloghttps://qwen.ai/blog?id=qwen3tts-0115

:laptop: GitHubhttps://github.com/QwenLM/Qwen3-TTS


3 seconds of audio. 3 lines of code. Unlimited voices.

RIP subscription TTS.

17 Likes

thanks you great benefit

1 Like

Bro, how to install it on Mac air m series? I tried importing this in ComfyUI by your given repo, but I tried generating it, and it did not work. I have no experience with ComfyUI.

my requirement is Indic languages, but there is no support for it :frowning: