ποΈ Audio8-TTS-Realistic-Air is now on Pinokio β 1-click install
Edge0's lightweight 0.6B text-to-speech model runs fully on your own machine.
π§ What it does
π Speaks 23 languages: English, Chinese, Cantonese, Japanese, Korean, Spanish, French, German, Italian, Portuguese,
Russian, Arabic, Hindi and more. It picks up the language from your text, so there's nothing to select.
ποΈ Clones a voice from a 3β15 second clip and its exact transcript. No training needed.
π² Has two built-in example voices, or choose "No reference" to get a random new voice.
π Outputs clean 44.1 kHz audio.
π₯οΈ Comes with a web UI where you can adjust temperature, top-p, top-k, length and seed.
π Includes a local API at /generate for your own apps and scripts.
π‘ Tips
βοΈ Short passages work best. One generation tops out at about 47 seconds.
π€ Use a clean recording of a single speaker, and type its transcript exactly.
π Before you install
π» An NVIDIA GPU is recommended, but the model is small, so it doesn't need much VRAM. CPU also works, but slower.
πΎ The weights are about 2.6 GB and download automatically on first start.
π Licensed Apache-2.0. Commercial use is allowed.
π Please only clone voices you have permission to use.
π€ https://huggingface.co/Edge0/Audio8-TTS-Realistic-Air
π Install β Start β Generate
ποΈ Audio8-TTS-Realistic-Air is now on Pinokio β 1-click install
Edge0's lightweight 0.6B text-to-speech model runs fully on your own machine.
π§ What it does
π Speaks 23 languages: English, Chinese, Cantonese, Japanese, Korean, Spanish, French, German, Italian, Portuguese,
Russian, Arabic, Hindi and more. It picks up the language from your text, so there's nothing to select.
ποΈ Clones a voice from a 3β15 second clip and its exact transcript. No training needed.
π² Has two built-in example voices, or choose "No reference" to get a random new voice.
π Outputs clean 44.1 kHz audio.
π₯οΈ Comes with a web UI where you can adjust temperature, top-p, top-k, length and seed.
π Includes a local API at /generate for your own apps and scripts.
π‘ Tips
βοΈ Short passages work best. One generation tops out at about 47 seconds.
π€ Use a clean recording of a single speaker, and type its transcript exactly.
π Before you install
π» An NVIDIA GPU is recommended, but the model is small, so it doesn't need much VRAM. CPU also works, but slower.
πΎ The weights are about 2.6 GB and download automatically on first start.
π Licensed Apache-2.0. Commercial use is allowed.
π Please only clone voices you have permission to use.
π€ https://huggingface.co/Edge0/Audio8-TTS-Realistic-Air
π Install β Start β Generate