gradium/phonon-31m-en-audiocpp

phonon-31m English for audio.cpp

phonon-31m-en converted for audio.cpp's gradium_phonon family. It is the same model; only the file format differs.

These files need an audio.cpp build with the gradium_phonon family, which is not merged yet. They do not load in llama.cpp: for llama.cpp, use gradium/phonon-31m-en-llamacpp.

For the Phonon runtime (command line, Python, browser, iOS, Android), use gradium/phonon-31m-en.

Files

File	What it holds
phonon-31m-en-q8_0.gguf	The model, with its config and tokenizer. The transformer's linear layers are 8-bit; the rest, the Mimi decoder included, keeps the original precision.
default-voice.safetensors	The default voice, the same file as in phonon-31m-en.

Quickstart

From an audio.cpp checkout:

python3 tools/model_manager_v2.py install gradium_phonon_31m_en_q8_0 --models-root models

audiocpp_cli --task tts --family gradium_phonon --model models/phonon-31m-en \
  --backend cpu --text "Hello from Phonon." --out speech.wav

--backend selects another audio.cpp backend, such as metal, and --mode streaming streams the audio.

Good to know

• Voices: without --voice-id, the default voice is used. More voices are in gradium/phonon-voices: copy one into the model folder as embeddings/<name>.safetensors and pass --voice-id <name>. The model has no voice cloning from audio.
• Text: audio.cpp sends the text to the tokenizer as written. Write numbers, symbols and abbreviations out in words. The Phonon runtime does this for you with --lang en.
• Sampling: temperature 0.3, the Phonon runtime's default.

License

Apache 2.0, like phonon-31m-en. Phonon builds on Pocket TTS and the Mimi codec, developed by Kyutai.

All Phonon models · Runtime on GitHub