Spaces:
Running
Running
| title: Kitten TTS Demo | |
| emoji: 😻 | |
| colorFrom: indigo | |
| colorTo: green | |
| sdk: gradio | |
| sdk_version: 6.6.0 | |
| app_file: app.py | |
| pinned: false | |
| python_version: "3.12" | |
| short_description: Kitten TTS 2 plus the open-source 0.8 models | |
| # Kitten TTS Demo | |
| Type text, pick a model and a voice, and hear it. | |
| - **Kitten TTS 2**, KittenML's newest and most expressive model: voices in ten languages (more in preview), | |
| emotion tags such as `[joyful]`, sounds such as `<laugh>` and `(((emphasis)))`. It comes from the | |
| [KittenML API](https://platform.kittenml.com), so it needs no GPU. | |
| - **Kitten TTS 0.8 Mini, Micro and Nano**, the open-source models (Apache 2.0). They are loaded when the Space | |
| starts and run on its CPU. | |
| The Space needs one secret, set under **Settings → Variables and secrets**: | |
| - `KITTENML_API_KEY`: a KittenML API key, used for Kitten TTS 2. The 0.8 models work without it. | |
| The Kitten TTS 2 voice list is read from the API, so new voices appear without a code change. Like the demo on | |
| kittenml.com, each visitor gets 4 requests a minute (12 per network) and up to 500 characters per request, and | |
| Kitten TTS 2 adds an audio allowance of 5 minutes every six hours (20 per network). | |
| The open-source models are on [GitHub](https://github.com/KittenML/KittenTTS) and | |
| [Hugging Face](https://huggingface.co/KittenML). | |