Twitch VOD Transcript Extractor

Extract transcripts from Twitch VODs and clips. Give it a Twitch video URL and get back timestamped text — ideal for turning stream archives into searchable content, generating captions for highlights, or analyzing gameplay commentary at scale.

Works with any public Twitch VOD or clip. No login required.

What it does

1. You provide a Twitch VOD or clip URL.
2. The actor extracts the spoken transcript with precise timestamps.
3. You get back structured text in JSON, plain text, SRT, or WebVTT format.

Twitch VODs can run for hours — a full stream session, a tournament broadcast, or a marathon charity event. This actor handles that scale and returns the same flat, timestamped structure regardless of duration.

Key features

• VOD & clip support — works with full VOD archives (/videos/<id>) and short clips (clips.twitch.tv/<slug>).
• Long-duration handling — Twitch streams often exceed two hours. The actor is designed to handle extended video without truncation.
• Timestamped segments — every segment includes start and end times, making it possible to jump to any moment in a long VOD by text search.
• 4 output formats — JSON with timestamps, plain text, SRT subtitles, or WebVTT for downstream video editors and players.
• Multi-language — uses any available captions; speech-to-text fallback covers 14 languages.
• No login required.

Supported languages

When Twitch provides auto-generated or manual caption tracks, the actor uses them. When captions are unavailable, speech-to-text fallback supports:

Code	Language
en	English
es	Spanish
fr	French
de	German
pt	Portuguese
it	Italian
nl	Dutch
ja	Japanese
ko	Korean
zh	Chinese
ar	Arabic
ru	Russian
tr	Turkish
hi	Hindi

Input

Field	Type	Default	Description
url	string	(required)	Twitch VOD or clip URL
language	string	"en"	Preferred language (ISO 639-1)
format	string	"json"	"json", "text", "srt", or "vtt"

Supported URLs

https://www.twitch.tv/videos/1234567890
https://clips.twitch.tv/FunnyClipSlugHere

Output example

When format is json, each dataset row looks like this:

{
    "url": "https://www.twitch.tv/videos/1234567890",
    "platform": "twitch",
    "language": "en",
    "duration_sec": 7452,
    "segments": [
        { "start": 0.0, "end": 5.2, "text": "What is up everyone, welcome back to the stream." },
        { "start": 5.2, "end": 11.0, "text": "Today we are doing a full ranked grind, no breaks until Diamond." }
    ],
    "full_text": "What is up everyone, welcome back to the stream. Today we are doing a full ranked grind, no breaks until Diamond."
}

Note: duration_sec for long VODs can be in the thousands. The segments array covers the full duration.

When format is text, srt, or vtt, the formatted transcript is written to the actor's key-value store under the OUTPUT key.

When extraction fails, the dataset row contains a clear error:

{
    "error": "NOT_FOUND",
    "message": "The video is private, removed, or has no extractable transcript."
}

Error codes

Code	Meaning
INVALID_FORMAT	format was not one of json/text/srt/vtt
BAD_REQUEST	The URL was malformed or rejected by the transcript operation
UNSUPPORTED_PLATFORM	No transcript operation serves this URL's platform
UNAUTHORIZED	The actor's ScriptBase key is missing, invalid, or revoked
INSUFFICIENT_CREDITS	The ScriptBase account behind this actor is out of credits
NOT_FOUND	The video is private, removed, or has no extractable transcript
UPSTREAM_FAILED	Every transcript source failed -- retry shortly
RATE_LIMITED	Too many requests -- retry after a short delay
TIMEOUT	Extraction took longer than the actor's poll budget
INTERNAL_ERROR	Unexpected error during extraction

Use cases

• Searchable stream archives — full VODs are hard to skim. Extract the transcript and index it so viewers, editors, or community managers can jump to any topic by keyword.
• Clip & highlight captioning — export SRT or WebVTT for a clip to add burned-in or soft subtitles before posting to YouTube Shorts, TikTok, or Instagram Reels.
• Esports & gameplay analysis — capture commentator analysis, callouts, and strategy discussion from tournament broadcasts. Feed text into analysis pipelines or LLMs.
• Moderation & compliance — scan transcripts for policy violations, prohibited phrases, or brand-safety issues without watching hours of footage.
• Content repurposing — turn a streamer's best moments or tutorial segments into written guides, forum posts, or wiki entries.
• Audience research — aggregate transcripts across multiple streamers in a category to understand what language, games, and topics drive engagement.

Integrations

Use this actor through the Apify API or integrate it into your content pipeline:

• Python — apify_client.actor("your-actor-id").call(run_input={"url": "https://www.twitch.tv/videos/1234567890"})
• JavaScript — await client.actor("your-actor-id").call({url: "https://clips.twitch.tv/FunnyClipSlugHere"})
• Zapier / Make / n8n — trigger on a new VOD URL in a spreadsheet and automatically generate a transcript for your content library.
• Webhooks — get notified when a long VOD transcript finishes, triggering summarization or archival workflows.

FAQ

Does this need a Twitch account or API key? No. The actor works with any publicly accessible VOD or clip without credentials.

My VOD is three hours long — will it still work? Yes. The actor is built to handle long-duration Twitch content. Note that extraction time scales with the length of the video; set a generous actor timeout for very long streams.

Are subscriber-only VODs supported? No. Only public VODs and clips accessible without a Twitch account are supported.

What if a clip has no captions? The actor applies speech-to-text extraction. See the supported languages table for coverage.

Can I extract transcripts from multiple VODs in one batch? Use the Apify API in a loop or build a Make/n8n scenario that iterates over a list of VOD URLs.

How do I find a VOD URL if I only have a channel name? Browse the streamer's Videos tab on Twitch, open a VOD, and copy the URL from your browser's address bar.

What does it cost? Pricing is per transcribed minute (1 minute = 1 charge unit, minimum 1), so you pay only for the audio actually processed — and failed extractions cost nothing. The exact per-minute price is shown on this actor's Apify Store page.