Twitter (X) Space Scraper

X Space Collector — Metadata, Hosts, Replay & Transcripts

Collect public X (Twitter) Spaces without an account, cookies or API keys. Start from Space links, IDs or announcement posts — or let the actor discover Spaces from a keyword, hashtag or host handle. Every run returns clean, omit-empty records for the Space, its people, its replay audio and, when requested, a timed transcript.

Three ways to find Spaces

Mode	Input	What happens
Exact Spaces	spaceUrls	Reads a Space link, a bare Space ID, or the post that announces a Space. The Space card is resolved automatically.
Keyword / hashtag	searchQueries	Reads recent public posts matching the query and resolves the Spaces they announce.
Host	hostUsernames	Reads a host's recent public posts (and their search results) and resolves the Spaces they announce.

All three modes can run in one go. Discovered Spaces are capped by maxSpaces; pasted links and IDs are always collected. When several inputs resolve to the same Space (for example its link and its announcement post), it is collected once. spaceState can keep only live, scheduled, ended or any Spaces.

What you get

Each run writes rows with a type:

type	Meaning
space	One Space with metadata, counts, host and participants
space participant	One row per host, speaker or listener
video transcript	The transcript summary with timed segments
space transcript segment	One row per timed segment (optional)
status	An honest problem row (invalid link, deleted Space, rate limit, unavailable replay)

Space row

• Identity: id, url, title, state, contentType
• Times: createdAt, scheduledStart, startedAt, endedAt, durationSeconds, updatedAt, replayStartSeconds
• Audience: listenersLive, listenersReplay, replayAvailable
• People: creator (full public profile), hostId, hostUsername, hostName, participants[], participantCounts
• Space settings: topics, topicIds, mentionedUserIds, isMuted, isLocked, isEmployeeOnly, disallowJoin, noIncognito, conversationControls, maxAdminCapacity, maxGuestSessions, narrowCastSpaceType
• Replay: replayAudioUrl, replayAudioFileUrl, replayAudioFileBytes, replayAudioContentType, replayAudioDurationSeconds
• Source: sourceTweetId, sourceTweetUrl when you started from an announcement post
• mediaKey, collectedAt

Participant row

spaceId, spaceUrl, spaceTitle, participantIndex, role, userId, username, name, bio, location, website, followers, following, postCount, profileImageUrl, bannerUrl, isBlueVerified, verified, verifiedType, periscopeUserId, isMutedByAdmin, isMutedByGuest, collectedAt.

Transcript row

spaceId, spaceUrl, title, language, transcriptionMethod (captions or ai), modelUsed, segmentCount, fullTranscript, segments[] (index, start, end, text), collectedAt.

Empty fields are omitted at every depth. There are no null values in the dataset.

Input

Field	Type	Default	Description
spaceUrls	array	–	Space links, Space IDs, or announcement post links
searchQueries	array	–	Keywords or hashtags for Space discovery
hostUsernames	array	–	Handles whose recent posts announce Spaces
maxSpaces	integer	20	Cap for discovered Spaces (1–200)
spaceState	string	any	any, live, upcoming or ended
includeParticipants	boolean	true	Participant list on the Space row + participant rows
includeParticipantProfiles	boolean	true	Enrich participant stubs with public profile details
maxParticipantProfiles	integer	50	Profile lookups per Space (0 = all)
includeReplayAudio	boolean	false	Replay HLS link + downloadable audio file
includeTranscript	boolean	false	Produce a transcript from captions or Whisper
transcriptSegmentsAsRows	boolean	false	Emit one row per transcript segment
transcriptionMethod	string	auto	auto, captions or ai
whisperModel	string	base	tiny, base, small, medium or large-v3
proxyConfiguration	object	–	Optional; automatic fallback is built in

At least one of spaceUrls, searchQueries or hostUsernames is needed.

Examples

Metadata and participants:

{
  "spaceUrls": ["https://x.com/i/spaces/1YpKkZgnrDXxj"],
  "includeParticipants": true
}

Discover by keyword, keep only ended Spaces:

{
  "searchQueries": ["#Space", "NASA"],
  "spaceState": "ended",
  "maxSpaces": 10
}

Discover from hosts:

{
  "hostUsernames": ["NASA", "@SpaceX"],
  "maxSpaces": 5
}

Replay audio with a transcript:

{
  "spaceUrls": ["1YpKkZgnrDXxj"],
  "includeReplayAudio": true,
  "includeTranscript": true,
  "transcriptionMethod": "auto",
  "transcriptSegmentsAsRows": true
}

Example Space record

{
  "type": "space",
  "id": "1YpKkZgnrDXxj",
  "url": "https://x.com/i/spaces/1YpKkZgnrDXxj",
  "title": "Meet NASA's New Mission Control Flight Directors #AskNASA",
  "state": "Ended",
  "createdAt": "2022-07-19T15:25:30+00:00",
  "startedAt": "2022-07-19T15:59:30+00:00",
  "endedAt": "2022-07-19T16:45:16+00:00",
  "durationSeconds": 2746.0,
  "listenersLive": 47951,
  "listenersReplay": 112570,
  "replayAvailable": true,
  "hostUsername": "NASA",
  "participantCounts": {"total": 10, "admin": 1, "speaker": 9},
  "topics": ["NASA", "Science", "Space"],
  "collectedAt": "2026-11-07T09:15:00+00:00"
}

How discovery works

Spaces are discovered from public posts, not from a private search index. The actor reads recent matching posts through X's public guest surfaces and resolves each post's Space card. This keeps the actor cookieless. Consequences:

• Discovery looks at recent public posts, not the entire archive.
• A Space with no public announcement post cannot be discovered by keyword or host. Paste its link or ID instead.
• Guest search is the least reliable public surface; a query can return fewer posts at busy times. The actor retries and falls back to the host's own timeline.
• X's signed-in Space search and topic browser are not available to guest callers. They are deliberately not faked.

FAQ

Do I need an X account, cookies or an API key? No. The actor is fully cookieless and reads only public Spaces, public posts and public profiles.

Does it work with private or deleted Spaces? No. They produce a clear status row.

Why is replay audio sometimes unavailable? Live Spaces have no replay yet, and some hosts disable replay. You get a status row explaining which case applies.

How is the replay file produced? The public replay stream is downloaded and combined into one playable audio file, then stored with a stable download link. Very large replays (above roughly 250 MB) may exceed the file cap — you still get the streaming replay link.

How do transcripts work? With auto, native captions are used when the replay has them; otherwise Whisper AI runs over the audio. Captions are fast. AI time grows with the Space length, and long Spaces are split into 10-minute chunks to bound memory.

Which Whisper model should I use? base is the best balance on the default 1 GB. Larger models need more actor memory. The actor downgrades safely instead of crashing, and reports the model it actually used in modelUsed.

How many rows will a run produce? A Space with participants and a full transcript can produce several hundred rows. Longer Spaces and larger participant lists produce more. Discovery is capped by maxSpaces; pasted Spaces are not capped.

Can I schedule it? Yes. Use searchQueries or hostUsernames with a schedule to pick up newly announced Spaces, or run it after a recurring show with the host's handle.

Is this affiliated with X? No. It is a third-party actor that reads publicly visible content.

Limits

• Public Spaces only; private, deleted or unavailable Spaces return a status row.
• Keyword/hashtag/host discovery covers recent public posts, not the full archive.
• Signed-in Space search and the topic browser are not reachable without credentials and are not supported.
• Replay audio and transcripts require a finished Space with replay enabled.
• AI transcription is CPU-bound; on 1 GB the actor runs tiny/base and may downgrade a larger requested model.
• Participant profile enrichment stops at maxParticipantProfiles per Space.

Integrations

Use Apify's built-in integrations to send Space summaries, participant lists or transcripts to Google Sheets, Airtable, Slack, a webhook or your own API.