encodec_24khz
by facebook
High-fidelity neural audio codec for compression and embeddings
facebook/encodec_24khzmixpeek://audio_extractor@v1/facebook_encodec_24k_v1Overview
EnCodec is a neural audio codec that compresses audio at various bitrates while maintaining high quality. Its encoder produces rich intermediate representations that serve as effective audio embeddings.
On Mixpeek, EnCodec provides audio embeddings that capture tonal, rhythmic, and structural audio features — useful for music similarity, audio fingerprinting, and content-based audio retrieval.
Architecture
Encoder-decoder architecture with a quantized bottleneck using Residual Vector Quantization (RVQ). Encoder: 1D causal convolutions. Decoder: mirrored architecture. Trained with reconstruction, perceptual, and adversarial losses.
Mixpeek SDK Integration
import { Mixpeek } from "mixpeek";
const mx = new Mixpeek({ apiKey: "API_KEY" });
await mx.collections.ingest({
collection_id: "my-collection",
source: { url: "https://example.com/song.mp3" },
feature_extractors: [{
name: "audio_embedding",
version: "v1",
params: {
model_id: "facebook/encodec_24khz"
}
}]
});Capabilities
- Variable bitrate audio compression (1.5-24 kbps)
- Rich intermediate audio representations
- High-fidelity audio reconstruction
- Supports mono and stereo at 24kHz
Use Cases on Mixpeek
Specification
Research Paper
High Fidelity Neural Audio Compression
arxiv.orgBuild a pipeline with encodec_24khz
Add this model to a processing pipeline alongside other extractors. Combine with retrieval stages for end-to-end search.
Open Pipeline Builder