Pocket TTS Greek
huggingface.co- Category
- AI Tools
- Rank
- No. 2480Tools index
Previous survey · No. 2382 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Design & Media
- Date
About
Pocket TTS Greek is a Greek and Cypriot Greek text-to-speech model for Pocket TTS with about 110 million parameters. It runs roughly three times faster than real time on a laptop CPU with streaming output. It was trained on about 2,400 hours of speech, includes eight curated voices (two Cypriot), and supports voice cloning from a 10 to 15 second sample.
What it does
The site describes a speech synthesis model for Greek and Cypriot Greek, built on Kyutai's Pocket TTS. It says the model copies a voice from a short clean recording, ships eight named voices, and streams audio faster than real time on a laptop processor. The card also lists weak spots, such as very short replies.
Stated on the product site
- Languages
- Greek and Cypriot Greek speech output
- License
- CC BY 4.0 weights, commercial use allowed with attribution
- Runtime
- Runs on a CPU with streaming output, per the site
- Model size
- About 110 million parameters
- Interfaces
- Command-line generation through uvx and a Python loading interface
- Voice cloning
- Clones from a 10 to 15 second sample, plus eight bundled voices
- Base model
- Fine-tuned from Kyutai's Pocket TTS
Not stated on the site
- All quality numbers on the page come from the authors' own 80-sentence test set; the page shows no outside evaluation.
- The site says Cypriot output was judged by ear only, so its accuracy has no measured figure on the page.
- Memory use and the delay before the first audio arrives are not stated.
- The page does not say how the model performs on GPUs or on hardware other than a laptop CPU.
Written from the product site at huggingface.co.
Tags
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.