Phonon-2

Phonon-2 is a single file. Phonon-1 remains available beside it.

The Phonon-1 spec

Phonon-2 averages 5.21 % word error on the seven public test sets of the Open ASR Leaderboard, scored with the board’s own code on the full test sets. The table sets it beside its full-precision teacher, Parakeet TDT 0.6B v3, which averages 4.96 in a 2,508 MB download, and six other open models from 178 MB to about 8 GB.

One 164 MB file runs on every surface, and the fast path on each keeps the accuracy of the exact one.

Detta, the dictation app for the Mac, runs Phonon-2 in any text field.

$ pip install fermion-research
$ phonon transcribe meeting.wav
$ phonon serve --port 8010
$ phonon listen
$ fermion transcribe phonon-2 meeting.wav

$ pip install mlx mlx-audio mlx-lm soundfile scipy zstandard

$ pip install --no-deps torch --index-url https://download.pytorch.org/whl/cpu
$ pip install fermion-research torch safetensors soundfile scipy zstandard

$ docker run --rm -v "$PWD":/audio -v phonon-cache:/home/phonon/.cache ghcr.io/fermionresearch/phonon-cpu:2.0.9 transcribe phonon-2 /audio/recording.wav

$ docker run --rm --gpus all -v "$PWD":/audio -v phonon-cache:/home/phonon/.cache ghcr.io/fermionresearch/phonon-cuda:1.0.8 transcribe phonon-2 /audio/recording.wav

Weights on Hugging FaceDocumentationfermion-research on PyPIEngines on GitHub

Phonon-2 Core ML runs the model on the Apple Neural Engine with the accuracy of the reference engine. One 332 MB package holds the encoder for every utterance up to 35 seconds, beside a 13 MB decoder, and on an M5 MacBook Air it transcribes an hour of speech in 6 seconds, 606 times realtime. Detta runs it in any text field, and the package runs it from Swift or Python.

$ python3.12 -m venv .venv && source .venv/bin/activate
$ pip install huggingface_hub
$ hf download FermionResearch/Phonon-2-CoreML --local-dir Phonon-2-CoreML

$ git clone --branch 1.1.1 https://github.com/fermionresearch/phonon-coreml
$ (cd phonon-coreml && swift build -c release)
$ phonon-coreml/.build/release/phonon-coreml-cli Phonon-2-CoreML recording.m4a --words

import PhononCoreML
let transcriber = try Transcriber(bundle: URL(fileURLWithPath: "Phonon-2-CoreML"))
let result = try transcriber.transcribe(url: URL(fileURLWithPath: "recording.wav"))
print(result.text)
for word in result.words { print(word.start, word.end, word.text) }

$ pip install coremltools numpy soundfile scipy
$ python Phonon-2-CoreML/phonon_coreml.py Phonon-2-CoreML recording.wav

The Swift runner reads wav, m4a, mp3, aiff, caf and flac at any sample rate and length, and the Python runner reads wav and flac at any sample rate. Both cut long recordings at pauses. The first run prepares the package for the Neural Engine once, which can take a minute or two, and later loads take under a second.

Core ML package on Hugging FaceSwift package on GitHub