Clone Any Voice in 3 Seconds — Free Forever
Alibaba just killed the $22/month voice cloning industry.
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
Clone any voice from 3 seconds of audio
Create voices by describing them in plain English
Control emotion, tone, speed
10 languages out the box
Runs 100% offline after setup
Apache 2.0 — use it for anything
Try It Now (Literally 0 Install)
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.
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 CLI — https://github.com/daliusd/qtts
./qtts.py "Hello world" -s Ryan -l English
REST API + Docker — https://github.com/ValyrianTech/Qwen3-TTS_server
docker run --gpus all -p 7860:7860 qwen3-tts_server
ComfyUI Nodes — https://github.com/DarioFT/ComfyUI-Qwen3-TTS
Links
Models — https://huggingface.co/collections/Qwen/qwen3-tts
Demo — https://huggingface.co/spaces/Qwen/Qwen3-TTS
Blog — https://qwen.ai/blog?id=qwen3tts-0115
GitHub — https://github.com/QwenLM/Qwen3-TTS
3 seconds of audio. 3 lines of code. Unlimited voices.
RIP subscription TTS.
!