Aside's speech models for Windows and Linux

ONNX files the Aside desktop app downloads on Windows and Linux, built by desktop/crates/aside-asr/parity/build_models.sh so they match the Mac app (FluidAudio 0.17.4) as closely as possible:

parakeet-tdt-0.6b-v2/encoder-model.onnx: NVIDIA Parakeet TDT 0.6B v2 encoder, 8-bit weights.
parakeet-ultra/encoder-model.onnx: moondream's Parakeet Ultra (from NVIDIA Parakeet TDT 0.6B v3) encoder, 8-bit weights.
speaker-diarization/*: FluidInference's speaker-diarization-coreml (pyannote segmentation, WeSpeaker, VBx PLDA), converted to ONNX.
parakeet-ctc-110m/ctc.onnx: NVIDIA Parakeet CTC 110M.

Credits: NVIDIA (Parakeet, CC-BY-4.0), moondream (Parakeet Ultra), pyannote and WeSpeaker, FluidInference, and the ONNX exports by istupakov and Masterx.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support