Avatar Lip Sync 1

Core ML build of LAM Audio2Expression (through wav2arkit_cpu) that Enconvo for Mac uses to move an Avatar's lips with its voice.

File	Size	Input	Output
`Lam.mlmodelc`	72 MB	`audio_waveform`, Float16 `[1, 32000]`: 2 s of 16 kHz mono speech	`blend`, Float16 `[1, 60, 52]`: 30 fps ARKit blendshape weights

macOS 14 or later, Apple silicon. Load it with compute units CPU and GPU: one 2 s window takes about 11 ms on an M1 Max. On the Neural Engine its output drifts from the original model.
Weights: 6-bit k-means palettized, 16-bit float compute.

License

Apache License 2.0. See LICENSE and NOTICE, which must ship with the model.

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

Base model

Finetuned

(1)

this model