Public registry snapshot: 2026-10-03. Enable JavaScript or follow Public model JSON for current registry data.
Discover the actual models and recipes.
curl --fail --silent --show-error \
https://core.modeluse.com/v1/modelsTHE MODEL LIBRARY
13 of 13 sources
openai
APACHE-2.0Speech recognition
Compact transcription model for speech and bounded audio experiments.
Private CPU transcription test preview
KimberleyJSN
MITSong separation
Vocal and instrumental separation model used in the Voice program.
Source catalog entry · audio inference unavailable here
lj1995
MITVoice conversion
Reference components for the RVC voice conversion pipeline.
Source catalog entry · audio inference unavailable here
Vyvo-Research
UNKNOWNAudio quality
Synthetic-speech naturalness prediction; distinct from NISQA transmitted-speech quality.
Source catalog entry · audio inference unavailable here
Vyvo-Research
UNKNOWNAudio quality
Transmitted-speech quality prediction for overall, noise, discontinuity, coloration and loudness dimensions.
Source catalog entry · audio inference unavailable here
microsoft
MS-PLAudio quality
Contrastive clean-quality prediction using published PAM prompts and MS-CLAP 2023 weights.
Source catalog entry · audio inference unavailable here
vkuzgn
MITAudio quality
No-reference natural-speech quality prediction; distinct from the synthetic-domain checkpoint.
Source catalog entry · audio inference unavailable here
niobures
UNKNOWNAudio quality
No-reference quality prediction using the synthetic/voice-conversion domain checkpoint.
Source catalog entry · audio inference unavailable here
unilight
MITAudio quality
Mixed-domain mean-opinion-score prediction using a WavLM Large checkpoint.
Source catalog entry · audio inference unavailable here
Blinorot
MITAudio quality
Speech naturalness prediction using a TorchScript UTMOS 2022 implementation.
Source catalog entry · audio inference unavailable here
sarulab-speech
MITAudio quality
Speech naturalness prediction using UTMOSv2.
Source catalog entry · audio inference unavailable here
mustafa-ozan-duman
APACHE-2.0Audio quality
English speech mean-opinion-score prediction using WavLM features and a transformer.
Source catalog entry · audio inference unavailable here
laion
APACHE-2.0Audio enrichment
Fixed relative acoustic-content scores and normalized private CLAP embeddings. No calibrated confidence or quality guarantee.
Source catalog entry · audio inference unavailable here
PIPELINES
Compare audio quality using specialist predictors and separate listening review.
Setup only · no public submission
Inputs → API → productionAttach speech-quality features and speaker/audio embeddings to catalog samples for comparison and similarity discovery.
Internal integration · no public submission
Inputs → API → productionValidate and decode your own WAV, transcribe with pinned Whisper Tiny, and retrieve a verified private result.
Qualified transcription preview · existing access required
Upload → quote → private resultFour fixed relative audio/text scores and a normalized private CLAP embedding for your own clip.
Audio enrichment unavailable · inspect contract
Upload → quote → private resultPrivate CPU transcription preview enabled in the registry. Sign-in and operator-issued development units are required.
Read the limitsBEFORE YOUR FIRST RUN
Inspect the live recipe, its input contract and explicit availability. A source task or dated catalog snapshot does not enable execution. Private CPU transcription is an operator-enabled test preview; check workspace access and development units first.
Manifest exports have no external compute charge; a customer execution price is not published here. The private CPU preview uses operator-issued noncash development audio-second units. These are separate from monetary balances. Read the actual task quote before submitting; funding alone does not grant audio execution.
Sign in here, then open My jobs. Jobs and private outputs retain their Core identities when the website changes.
Your existing Core account and jobs remain. Browser-local drafts stay on the old origin and are not copied into a redirect. You can save an old request file locally in the original browser. Review your existing jobs before sending work again, especially after an interrupted submission. For help recovering an old saved request, contact us without sending the private file.