AI BriefingPocket TTS Releases Multilingual Model
Key point
Pocket TTS has released a multilingual model supporting 6 languages along with an ONNX export tool.
Details
Pocket TTS has released a multilingual model, supporting English, French, Spanish, German, Italian, and Portuguese.
Each language is provided as a separate model, and the author stated that they modified the export script for the multilingual version based on KevinAHM's ONNX exporter and VolgaGerm's Pocket TTS cpp optimization.
They also explained that they applied selective int8 quantization to some nodes to improve generation speed and stability.
Benchmark figures were also presented.
- AMD Ryzen 9 7950X: latency of about 30ms, up to 13x realtime
- Helio G99: latency of about 100ms, 2.5x realtime
Related code and usage examples are as follows.
- Pocket TTS ONNX exporter:
lookbe/pocket-tts-onnx-export - Pocket TTS Unity sample runner:
lookbe/pocket-tts-unity
The author added that an Android version is already available as well.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
StepAudio 3 Gen Achieves SOTA in TTS and Multimodal Audio Generation via RVQ-Based Discrete Generation