yamnet-tflite

Unmodified copies of Google's official MediaPipe model files, mirrored so that apps can pin them by commit (git submodule). Licence: Apache-2.0 (Google LLC). No conversion, quantisation or other change was made.

Files are loaded by native MediaPipe Tasks for Android (com.google.mediapipe:tasks-audio), never by web runtimes.

File Bytes SHA-256 Source Description
yamnet.tflite 4126810 4d8b4a53282dc83ef04e3e7dbc4fbc98082e34e44ed798e16c3a0cdd4c584faf https://storage.googleapis.com/mediapipe-models/audio_classifier/yamnet/float32/latest/yamnet.tflite YAMNet audio event classifier, float32, input 15600 samples (0.975 s at 16 kHz), 521 AudioSet classes (the label list is embedded in the file as yamnet_label_list.txt). Used by AudioClassifier (MediaPipe Tasks Audio).

Downloaded 2026-11-20.

Attribution (required): the 521 class names of YAMNet come from the AudioSet ontology, which Google publishes under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0); the AudioSet dataset itself is CC BY 4.0 (statement read on https://research.google.com/audioset/download.html on 2026-11-20). An app that ships this model and shows its class names must credit the AudioSet ontology (Google, CC BY-SA 4.0, https://research.google.com/audioset/) on its licences screen. The model file itself is Apache-2.0 (Google LLC).

Storage note: Hugging Face refuses binary files as plain git objects, so every file here is stored as an LFS/Xet object. A plain git clone without git-lfs yields pointer files of about 130 bytes; clone with git-lfs installed (or run git lfs pull) to get the real bytes. Verify with the SHA-256 above.

Downloads last month
8
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support