YouTube AI Transcriber — no captions needed

Gemini AI watches the actual video — audio and frames — and writes a verbatim transcript with timestamps and speaker labels. It works on videos that have no captions at all, in any spoken language, and exports ready-to-use SRT and WebVTT subtitle files plus JSON and plain text.

Caption scrapers can only copy captions that already exist. This actor transcribes the video itself.

What it does

• ✅ Transcribes videos without captions — AI listens to and watches the video
• ✅ Speaker labels — real names when identifiable on screen, otherwise Speaker 1/2/…
• ✅ Timestamps on every segment (MM:SS / HH:MM:SS)
• ✅ 4 output formats: SRT subtitles, WebVTT subtitles, JSON segments, plain text — each as a downloadable file
• ✅ Any spoken language, up to 150 minutes per video
• ✅ No API key needed — or bring your own Gemini key and pay 5× less

Pricing

Charged per minute of video (minimum 5 minutes per video), plus a $0.02 actor start fee.

Video length	No API key (bundled)	With your Gemini key
Short (≤5 min)	$0.75	$0.15
10 min	$1.50	$0.30
30 min	$4.50	$0.90
60 min	$9.00	$1.80

Bring your own key and save 5×: create a free key at aistudio.google.com/apikey and paste it into the geminiApiKey field (stored encrypted, used only for your run). Google token costs (a few cents per hour of video) are then billed to your Google account.

You are only charged when a transcript is actually delivered — failed or timed-out runs charge nothing beyond the start fee.

Input

Field	Type	Description
videoUrl	string (required)	Public YouTube video URL (watch, youtu.be, or Shorts link)
includeSpeakers	boolean	Label each segment with the speaker (default true)
outputFormats	array	Any of json, srt, vtt, text (default: all)
language	string	Optional language hint, e.g. "Estonian"
model	string	gemini-3.7-flash (default) or gemini-3.5-flash-lite
geminiApiKey	secret string	Optional — your Google AI Studio key for 5× cheaper runs

Output

One dataset item per run:

{
  "videoUrl": "https://www.youtube.com/watch?v=jNQXAC9IVRw",
  "videoTitle": "Me at the zoo",
  "durationSeconds": 19,
  "minutesCharged": 5,
  "language": "English",
  "segmentCount": 4,
  "transcript": "[00:01] Jawed Karim: Alright, so here we are in front of the, uh, elephants...",
  "segments": [
    { "start": "00:01", "end": "00:04", "speaker": "Jawed Karim", "text": "Alright, so here we are in front of the, uh, elephants." }
  ],
  "srtUrl": "https://api.apify.com/v2/key-value-stores/…/records/transcript-jNQXAC9IVRw.srt",
  "vttUrl": "…", "jsonUrl": "…", "textUrl": "…",
  "status": "completed"
}

The SRT/VTT/JSON/text files are saved to the run's key-value store with public download URLs — drop the SRT straight into a video editor or player.

Use it via API

curl -X POST "https://api.apify.com/v2/acts/andok~youtube-ai-transcriber/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"videoUrl": "https://www.youtube.com/watch?v=jNQXAC9IVRw"}'

FAQ

How is this different from YouTube transcript scrapers? Scrapers download captions that already exist. This actor has Gemini watch the video and transcribe it from the audio and visuals — so it works when captions are missing, auto-captions are garbage, or you need speaker labels.

What videos are supported? Public, finished YouTube videos up to 120 minutes (150 minutes with your own key). Private, unlisted, and live videos are not supported.

What languages? Any language Gemini understands — the transcript is written in the video's original spoken language.

How long does a run take? Typically well under the video's own duration — a 10-minute video usually transcribes in 1–3 minutes.

What if the run fails? You are never charged per-minute fees for a failed or undelivered transcript.