---
title: "Concepts"
description: "How the API works and how to choose between its parts."
---

# Concepts

How the API works and how to choose between its parts. Part of [AudD resources](https://audd.io/resources/index.md).

- [Audio formats and quality for music recognition](https://audd.io/resources/concepts/audio-format-and-quality.md): Which audio and video formats AudD accepts, the 10 MB standard-endpoint cap, and practical guidance on clip length, sample rate, channels, and compression for reliable matches.
- [Controlling enterprise endpoint cost](https://audd.io/resources/concepts/enterprise-cost-control.md): How the AudD enterprise endpoint bills per 12 seconds of audio, and how limit, every, skip, and skip_first_seconds trade cost against completeness.
- [From match to decision: designing allow, block, and review policies](https://audd.io/resources/concepts/match-to-decision.md): AudD reports which recordings matched; your platform decides what happens next. The policy inputs an AudD result gives you, how to compute matched seconds and the matched share of an upload from enterprise results, a worked decision table with code, and how to handle appeals, logging, repeat uploaders, and regional rules.
- [Interpreting the recognition score](https://audd.io/resources/concepts/score-thresholds.md): Every AudD match carries a confidence score. How to read it and how to set application-specific thresholds for auto-accept, blocking, and human review.
- [Public database vs your custom catalog](https://audd.io/resources/concepts/custom-vs-public-db.md): When AudD recognition matches against the public 160-million-song database versus a private custom catalog you upload, and how to tell the two apart in a response.
- [Retries: what's safe to retry and what consumes a request](https://audd.io/resources/concepts/retry-strategy.md): Which AudD API failures are safe to retry, which consume a request, and how to configure the SDK's retry budget without double-billing or duplicating catalog entries.
- [Standard, enterprise, or streams: how to choose](https://audd.io/resources/concepts/standard-vs-enterprise-vs-streams.md): Pick the right AudD recognition surface for your input: standard for a short clip, enterprise for long files, streams for continuous live audio.
- [The song_link (lis.tn) universal music link](https://audd.io/resources/concepts/lisn-song-link.md): Every AudD match includes a song_link on the lis.tn domain; how its ?thumb and ?provider conventions and the SDK helpers let you build cover-art and 'open in your service' UIs without requesting provider metadata.
- [Webhook callbacks vs longpoll for stream results](https://audd.io/resources/concepts/callback-vs-longpoll.md): Choose how to receive AudD stream recognition events: webhook callbacks when you can host a public HTTPS endpoint, longpoll when you can't.
- [What music recognition matches, and what it doesn't](https://audd.io/resources/concepts/what-recognition-matches.md): AudD identifies recordings, not songs. A case-by-case table of what to expect for re-uploads, noisy captures, remasters, live versions, covers, samples, sped-up audio, AI soundalikes, DJ mixes, and unreleased music, plus how to build a test set that measures it on your own audio.

## Sitemap

See the [resources sitemap](https://audd.io/resources/sitemap.md) for every guide and the [glossary](https://audd.io/resources/reference/glossary.md) for AudD terms.