Udio Scraper

Collect public tracks from Udio together with the text that made them: the complete generation prompt, the full lyrics, and the style tags. Those three fields are kept exactly as the creator wrote them, because they are usually the whole point of collecting Udio tracks and they are the first thing a lossy export drops.

Every row also carries the artwork, a direct audio link, like and play counts, duration, timestamps and the creator handle.

What you can collect

Mode	What it does	What you supply
Search the catalogue	Finds public tracks by keyword, in the order you choose	A keyword (optional) and an ordering
Specific tracks	Reads a list of tracks you already know	Track page addresses, or bare identifiers
Playlist	Reads every track in one or more public playlists, in playlist order	Playlist page addresses, or bare identifiers

Input

Field	Type	Default	Notes
mode	select	search	search, songIds or playlist
searchTerm	text	empty	Matched against titles, prompts and tags. Empty browses everything
sortBy	select	newest	newest, mostLiked or mostPlayed
trackIds	list	empty	One track per line, address or identifier
playlistIds	list	empty	One playlist per line, address or identifier
maxItems	number	20	Stop after this many tracks. 0 collects everything available
resumeFromRunId	text	empty	Continue one interrupted run without returning what it already saved
incrementalMode	checkbox	off	Return only what changed since the last run of the same selection
emitUnchanged	checkbox	off	Also return unchanged tracks, marked UNCHANGED (billed)
emitExpired	checkbox	off	Also return tracks that are gone, marked EXPIRED (billed)
stateKey	text	empty	Name a monitoring campaign, or share one across runs
mcpConnectors	list	empty	Optional MCP connectors to also send the results to. Empty skips the export
notionParentPageUrl	text	empty	Notion connector only: the page under which track pages are created
maxNotifyListings	number	50	How many tracks each connector receives. Does not affect the dataset
proxyConfiguration	proxy	Apify Proxy	The default works on every plan

Both list fields accept either shape, so you can paste addresses straight from your browser or feed in bare identifiers from an earlier run.

Resume and recurring updates

These are two different things.

Resume finishes ONE interrupted collection. Paste that run's ID (or its dataset ID) into resumeFromRunId and the tracks it already saved are skipped.

Incremental mode is for running the same selection again and again, daily say, to watch a playlist gain and lose tracks. The actor remembers the previous run of that selection and, from the second run on, returns only what changed. Each row then carries changeType, changedFields, firstSeenAt and lastSeenAt. Memory is kept per selection (mode, keyword, ordering and lookup list); set stateKey to name a campaign or to deliberately share one.

EXPIRED rows are only produced when a run scanned the whole selection. A run stopped by maxItems, a keyword search that hit the 200-track ceiling, a collection that could not be read, and a resumed run all skip that step and say so in the log: a partial scan must never report a live track as gone.

Output

Field	Description
trackId	Stable identifier for the track
url	Track page address
title	Track title
artist	Creator handle shown on the track
artistImageUrl	Creator avatar image
creatorId	Stable identifier for the creator
prompt	The full generation prompt, kept verbatim
lyrics	The complete lyrics, kept verbatim, with section markers intact
tags	Style tags attached to the track
userTags	Tags added by the creator
description	Free text description
attribution	Attribution text, when the creator set one
likeCount	Number of likes
playCount	Number of plays
durationSeconds	Length in seconds
createdAt	When the track was generated
publishedAt	When the track was made public
audioUrl	Direct link to the audio file
videoUrl	Direct link to the video, when one exists
artworkUrl	Cover artwork image
styleId	Identifier of the style used
styleSourceType	Where the style came from
styleSourceTrackId	Track the style was taken from, when applicable
parentTrackId	Track this one extends or remixes, when applicable
generationId	Identifier of the generation batch
generationStatus	Status reported for the generation
isFinished	Whether generation completed
isPublishable	Whether the track may be shared
collectionId	Playlist identifier, in playlist mode
collectionName	Playlist name, in playlist mode
collectionUrl	Playlist page address, in playlist mode
positionInCollection	Place in the playlist, in playlist mode
sourceMode	Which mode produced the row
scrapedAt	When the row was collected
raw	Every field exactly as the site supplied it, so nothing is lost
changeType	Incremental mode only: NEW, UPDATED, UNCHANGED, REAPPEARED or EXPIRED
changedFields	Incremental mode only: which columns moved since the last run
firstSeenAt	Incremental mode only: when this track was first collected
lastSeenAt	Incremental mode only: when it was last collected

Sample output

{
  "trackId": "00000000-0000-4000-8000-000000000000",
  "url": "https://www.udio.com/songs/00000000-0000-4000-8000-000000000000",
  "title": "Midnight Coffee Loop",
  "artist": "Jane Doe",
  "creatorId": "00000000-0000-4000-8000-000000000001",
  "prompt": "lofi, jazzy chill cafe vibes, warm tape saturation",
  "lyrics": "[Verse 1]\nRain on the window, cups going cold\n\n[Chorus]\nStay for one more song",
  "tags": ["chill", "jazz", "downtempo", "instrumental"],
  "userTags": ["study"],
  "description": "",
  "likeCount": 128,
  "playCount": 4210,
  "durationSeconds": 131.2,
  "createdAt": "2026-01-05T09:12:44+00:00",
  "publishedAt": "2026-01-05T09:15:02+00:00",
  "audioUrl": "https://example.com/audio/sample.mp3",
  "artworkUrl": "https://example.com/art/sample.png",
  "isFinished": true,
  "isPublishable": true,
  "sourceMode": "search",
  "scrapedAt": "2026-01-06T11:00:00+00:00"
}

Send results into your apps (MCP connectors)

Results can optionally be piped into apps you already use, through Model Context Protocol connectors, as a delivery step after the collection. Authorize a connector under Apify, Settings, API and Integrations, then select it in the mcpConnectors field of the input. For Notion, also set notionParentPageUrl to the page under which track pages should be created; maxNotifyListings caps how many tracks are written to each connector.

What arrives in the connector is a condensed, human readable summary per track: a title plus the key fields flattened to plain text. Nested objects collapse to their main value and long lists are trimmed to a few entries, so the prompt and lyrics may be shortened there. The complete record always stays in the Apify dataset, which this step never changes. Leaving mcpConnectors empty skips the export entirely.

Good to know

• A single catalogue search reaches at most 200 tracks, because that is as far as the site's own search view goes. To collect more, run several searches with narrower keywords instead of raising maxItems.
• Playlist mode uses the playlist's own track list, so it returns every member of the playlist rather than the shorter preview the site shows elsewhere.
• Tracks that have been removed or made private are reported in the run log and simply skipped, so a short result is never silent.
• A very rare or very long keyword can make the site's own search time out rather than answer. The run stops with a clear message when that happens, so it is never reported as a successful empty result. Try a broader keyword.
• maxItems defaults to 20 to keep a first run quick and cheap. Set it to 0 when you want everything.

Support

Found a track that does not come back correctly, or a field you need that is missing? Open an issue from the Actor page and include the input you used.