Snapchat Spotlight Video Transcript Extractor

Public Snapchat Spotlight clips carry spoken words that Snapchat does not always show as text in a reusable form. This actor returns a clean transcript for each clip — first trying the caption track Snapchat already published, then, when asked, generating one from the audio with speech recognition.

You can feed it exact Spotlight links, a creator's pinned Spotlight videos, a hashtag page, or a country's trending Spotlight feed. Every row also includes the creator, engagement counts, video timing and thumbnail so the transcript can be joined to the rest of your data.

What you get

• Published captions when available. Snapchat exposes a WebVTT caption file for many videos. The actor downloads it, parses the cue timings, and returns both segments and the concatenated fullText.
• Speech recognition as a fallback. For videos without published captions, auto mode transcribes the audio with Whisper. Three model sizes trade speed for accuracy.
• Four ways to choose videos. Paste links, pull a creator's pinned Spotlight highlights, browse a hashtag, or page through a country feed.
• Metadata alongside text. Creator handle and display name, avatar, title, description, embedded caption, keywords, hashtags, duration, upload date, resolution, and view/share/comment/like counts.
• Smaller datasets on demand. Turn off timed segments to keep fullText, wordCount and segmentCount without the per-cue array.
• Clean rows. Empty values are omitted instead of written as null, and each row carries recordType, sourceUrl and scrapedAt.

Modes

Mode	What it reads	Typical input
bySpotlightUrls	The exact videos you paste	https://www.snapchat.com/spotlight/<id>, https://www.snapchat.com/@user/spotlight/<id>, https://t.snapchat.com/<code>, or a bare snap ID
byUsername	A public creator's pinned Spotlight highlights	brentrivera, @brentrivera, https://www.snapchat.com/@brentrivera
byTopic	Videos listed on a hashtag page	comedy, #comedy, https://www.snapchat.com/tag/comedy
byFeed	A country's trending Spotlight feed, 25 videos per page	US, GB, PK

Discovery modes stop after maxVideosPerSource videos per handle, hashtag or feed. maxItems caps the total number of rows in the run.

Output fields

Field	Description
recordType	Always spotlightTranscript
snapId	Spotlight snap ID
sourceUrl	Canonical Snapchat page for the video
mode	bySpotlightUrls, byUsername, byTopic or byFeed
inputUrl	The link you supplied (paste-URL mode only)
discoveredVia	Handle, hashtag or country that surfaced the video (discovery modes only)
title	Snapchat's generated title, when present
description	Snapchat's generated description, when present
caption	Caption embedded in the video
username	Creator's Snapchat handle
creatorName	Creator's display name
creatorUrl	Creator profile link
creatorAvatarUrl	Creator avatar image link
videoUrl	Direct CDN video link (may expire)
thumbnailUrl	Thumbnail image link
durationSeconds	Video length in seconds
uploadedAt	Upload time, ISO 8601 UTC
width, height	Video dimensions in pixels
keywords	Keywords Snapchat associates with the video
hashtags	Hashtag entries from the keyword list
viewCount, shareCount, commentCount, likeCount	Public engagement counts at scrape time
transcriptSource	native for published captions, whisper for generated text
transcriptAvailable	Whether a usable transcript was produced
transcriptUrl	WebVTT caption file link, when one exists
language	en for published captions; detected or hinted code for generated transcripts
languageProbability	Language-detection confidence (0-1) for generated transcripts
whisperModel	Model size used for generated transcripts
segments	Timed cues: [{startTime, endTime, text}]
segmentCount	Number of cues
fullText	Whole transcript as one string
wordCount	Word count of fullText
scrapedAt	When the row was produced, ISO 8601 UTC
error	Explanation when fetching or transcription failed for that row

Input

Field	Type	Default	Description
mode	string	bySpotlightUrls	Which discovery path to use.
spotlightUrls	array	–	Links or IDs; required when mode is bySpotlightUrls.
usernames	array	–	Creator handles; used when mode is byUsername.
topics	array	–	Hashtags or topics; used when mode is byTopic.
country	string	US	Two-letter feed country; used when mode is byFeed.
transcriptionMethod	string	auto	auto, native or whisper.
whisperModel	string	base	tiny, base or small.
language	string	auto	Optional language hint for speech recognition.
includeSegments	boolean	true	Include the timed segments array.
maxItems	integer	100000	Maximum rows to push, including error rows.
maxVideosPerSource	integer	20	Videos collected per handle, hashtag or feed.

Legacy keys from the previous version (spotlightUrl, urls, startUrls, maxResults, method, model) are still accepted and mapped to the current fields when the new key is absent.

Paste a handful of links

{
  "mode": "bySpotlightUrls",
  "spotlightUrls": [
    "https://www.snapchat.com/spotlight/W7_EDlXWTBiXAEEniNoMPwAAYYmViYXF5amNyAZXOAULlAZXOAOzLAAAAAQ",
    "https://t.snapchat.com/example"
  ],
  "maxItems": 5
}

Transcribe a creator's pinned Spotlight videos

{
  "mode": "byUsername",
  "usernames": ["brentrivera"],
  "maxVideosPerSource": 10,
  "transcriptionMethod": "auto"
}

Browse a hashtag with a fixed transcript style

{
  "mode": "byTopic",
  "topics": ["comedy"],
  "maxVideosPerSource": 20,
  "transcriptionMethod": "whisper",
  "whisperModel": "small",
  "language": "en"
}

Page a country feed

{
  "mode": "byFeed",
  "country": "GB",
  "maxVideosPerSource": 50,
  "includeSegments": false
}

How it reads Snapchat

All requests are anonymous: no login, cookie, API key or proxy account is needed. The actor reads the public video page, its embedded Schema.org VideoObject data, the published WebVTT caption file, and the public Spotlight discovery surfaces (profile highlights, tag pages and the feed's public GraphQL query). When a page is unavailable, the transcript endpoint for that snap ID is used as a fallback.

Limits and notes

• Published captions exist only for some videos. In a mixed 34-video probe, 8 had native captions; the rest need auto or whisper if you want text.
• Native captions are returned with language: en, matching Snapchat's published caption behaviour.
• Creator mode uses pinned Spotlight highlights (usually around 16 videos), not a creator's complete posting history.
• Tag pages usually list 20-35 videos; feed pages return 25 videos per page and are paged until maxVideosPerSource is reached.
• CDN video and thumbnail links can expire. Re-run the actor or copy the files if you need durable media.
• A video with no speech, only music, or very low confidence returns transcriptAvailable: false plus an error instead of invented text.
• A removed, expired or region-blocked clip returns a row with transcriptAvailable: false and an explanatory error.
• Follower counts, sound/lens names and recommendation counts are not published on the clip page, so they are not returned.
• Comments, downloads, private profiles and login-gated content are outside this actor's scope.

FAQ

Do I need a Snapchat account?

No. The actor only reads public pages and public caption files.

What happens when a video has no published captions?

With auto, the actor downloads the video and runs speech recognition. With native, it returns transcriptAvailable: false and explains why.

Which Whisper model should I pick?

base is the usual choice. tiny is faster and cheaper for clear speech. small helps with accents and non-English audio.

Can I request transcripts in a particular language?

Set language to an ISO 639-1 code for speech recognition. Published captions are returned as Snapchat provides them.

Why is a transcript missing segments?

The video may have text but no caption file, or you may have turned off includeSegments. fullText, wordCount and segmentCount remain available.

Are results identical on every run?

Transcript text and metadata are stable, but engagement counters and CDN links change over time, and discovery modes can surface a slightly different set as feeds rotate.