Chatterbox TTS Multilingual | Audio Samples (23 Languages) & Voice Cloning
In this video, I take a detailed look at Chatterbox TTS, a multilingual text-to-speech system by Resemble AI that runs privacy aware on your local system. All 23 internationally provided tts models are under MIT license.
You’ll hear audio samples in 23 different languages, get an overview of the available models, and see how Chatterbox performs in real-world usage. I also test zero-shot voice cloning to evaluate how well the system can reproduce a voice from limited reference material.
The video covers:
Overview of Chatterbox TTS
Audio samples across 23 languages
Local installation and usage via Gradio
Zero-shot voice cloning test
Practical insights and lessons learned
00:00 Intro
02:25 First look
05:45 Command line | Test Python 3.12
07:38 Installing using Python 3.11
09:45 Running gradio local (synthesize audio)
15:18 Testing zero shot voice cloning
17:57 Start of audio samples with Arabic
18:05 Danish
18:13 German
18:23 Greek
18:32 English
18:41 Spanish
18:51 Finnish
19:01 French
19:09 Hebrew
19:19 Hindi
19:27 Italian
19:36 Japanese
19:45 Korean
19:54 Malay
20:04 Dutch
20:14 Norwegian
20:22 Polish
20:29 Portuguese
20:42 Russian
20:50 Swedish
21:00 Swahili
21:12 Turkish
21:23 Chinese
On problems with virtual machine usage:
echo 'export DNNL_MAX_CPU_ISA=AVX2' 2xLARGER_THEN_CHAR ~/.bashrc
source ~/.bashrc
Or add following line in python script while torch import:
torch.backends.mkldnn.enabled = False