codec.audio benchmarks neural audio codecs on noisy in-the-wild speech; MOSS tops
kastnerkyle · x · 2026-11-11
codec.audio is a new site for comparing neural audio codecs, evaluated on VoxLingua107's in-the-wild (noisy) dev set with per-language metrics. Trelis Research notes MOSS's codec is very good quality but large at 2B params. Charts show moss leading on speaker similarity, ahead of qwen3tfs, mimi, nanocodec, while smaller codecs like facodec and snac trail.
More from Multimodal
- MiniMax H3 renders 4K character portraits twice as fast as Seedream 5.5's 2K — heyronir · 2026-11-12
- PhyRubric: rubric-based benchmark diagnoses physical commonsense in text-to-video — Huiyi Chen · 2026-11-12
- Layer-and-mask workflow keeps local SD inpaint variants editable in PeerBrush — Deftroyy · 2026-11-12
- VAE round-trip test: MiniMax-H3-Image latent compression scores worst, below even SD 1.5 — ostrisai · 2026-11-12
- Adobe MAX keynote: how Drex Lee uses Firefly Boards to rehearse one-take films — rufusd · 2026-11-12
- Pomelli Lets Brands Lock AI Model Traits for Consistent Ad Campaigns — jacalulu · 2026-11-12