ReelSDK burns TikTok-style word-by-word captions into a video with one POST /v1/captions and returns 5 files, SRT and VTT included. At $0.03 per output minute, a 60-second TikTok is 3 cents and a 10-second clip half a cent.

TikTok-style word-by-word captions with one POST

Send a video URL and get it back with each word lit up as it is spoken; a 10-second clip takes about 13 seconds. New ReelSDK accounts get $0.50 of free credit, enough for 10 clips of 10 seconds.

Use casesUpdated

The bold preset on a 10-second sample, with the spelling hint.Input: input videocaptions.srtcaptions.vttcaptions.asstranscript.jsonone word at a timekaraoke

The request

POST /v1/captions

Burn TikTok-style word-by-word captions into a video with one POST and get SRT, VTT and a word-timed transcript back.

terminalTry it
curl "https://api.reelsdk.com/v1/captions?wait=60" \
  -H "Authorization: Bearer $REELSDK_KEY" \
  --json '{
    "video_url": "https://api.reelsdk.com/f/out/job_o5kpzf5opqvopafnvlpd/video.mp4",
    "style": "bold",
    "prompt": "octopus, gills, swims"
  }'
one-word.mjsTry it
const res = await fetch("https://api.reelsdk.com/v1/captions?wait=60", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.REELSDK_KEY}`, "Content-Type": "application/json" },
  body: JSON.stringify({
    "video_url": "https://api.reelsdk.com/f/out/job_o5kpzf5opqvopafnvlpd/video.mp4",
    "style": {
      "preset": "bold",
      "max_words": 1
    },
    "prompt": "octopus, gills, swims"
  }),
});
const { job } = await res.json();
karaoke.shTry it
curl "https://api.reelsdk.com/v1/captions?wait=60" \
  -H "Authorization: Bearer $REELSDK_KEY" \
  --json '{
    "video_url": "https://api.reelsdk.com/f/out/job_o5kpzf5opqvopafnvlpd/video.mp4",
    "style": "karaoke",
    "prompt": "octopus, gills, swims"
  }'
response.json
{
  "job": {
    "id": "job_3ijtm71e07rauxrgbw4h",
    "op": "captions",
    "status": "succeeded",
    "stage": null,
    "created_at": "2026-11-07T01:14:15.764Z",
    "finished_at": "2026-11-07T01:14:29.170Z",
    "outputs": {
      "video": "https://api.reelsdk.com/f/out/job_3ijtm71e07rauxrgbw4h/video.mp4?s=7c96e25c...",
      "transcript": "https://api.reelsdk.com/f/out/job_3ijtm71e07rauxrgbw4h/transcript.json?s=83325d5d...",
      "srt": "https://api.reelsdk.com/f/out/job_3ijtm71e07rauxrgbw4h/captions.srt?s=a8121355...",
      "vtt": "https://api.reelsdk.com/f/out/job_3ijtm71e07rauxrgbw4h/captions.vtt?s=fa1306d8...",
      "ass": "https://api.reelsdk.com/f/out/job_3ijtm71e07rauxrgbw4h/captions.ass?s=f3ded6d8..."
    },
    "result": null,
    "duration_s": 10.466,
    "error": null,
    "usage": {
      "output_seconds": 10.466,
      "audio_seconds": 10.453,
      "tts_characters": 0,
      "images": 0
    },
    "poll_url": "https://api.reelsdk.com/v1/jobs/job_3ijtm71e07rauxrgbw4h"
  }
}

How the word-by-word highlight works

POST /v1/captions sends the audio to Whisper large-v3-turbo on Cloudflare Workers AI, which returns a start and end time for every word. The captions are drawn from those timings, so the highlight moves word by word with the voice.

The style preset decides how the spoken word looks. bold, the default, puts 3 uppercase Montserrat words in the middle of the frame; the spoken word turns #FFE14D yellow, pops up to 116% and settles back to 106% within a fifth of a second. karaoke puts 5 Poppins words at the bottom and sweeps a #3DDCFF cyan fill across each word as it is said. minimal puts 7 smaller Poppins words (size 0.85) at the bottom and keeps the spoken word white, so it reads like plain subtitles.

At a glance

The three caption presets

Each preset is a full style; send style as an object to change any field.

The three caption presets
PresetTextWords at onceThe spoken word
bold (default)Montserrat, uppercase, middle3Turns #FFE14D yellow and grows to 116%
karaokePoppins, bottom5Fills with #3DDCFF cyan as it is said
minimalPoppins at size 0.85, bottom7Stays white, like the rest of the line

Source: ReelSDK captions docs, checked November 2026.

Spelling hints for names and jargon

The sample is a 10-second vertical slideshow narrated by our text-to-speech from a short script: "An octopus has three hearts and blue blood. Two of those hearts only pump blood to the gills. The third one stops beating every time it swims."

Without a hint, the transcriber heard the last word, "swims", as "swings". With "prompt": "octopus, gills, swims", all 27 words came back right. prompt takes up to 500 characters of names and terms. If you already have the words and their timings, send them in words (up to 50,000) and transcription is skipped.

One word at a time, or a full line

style takes a preset name or an object that overrides single fields. max_words sets how many words share the screen, from 1 to 8. With { "preset": "bold", "max_words": 1 } the video shows one word at a time, and its captions.ass has 27 dialogue lines, one per word.

The object also takes font (Montserrat, Anton, Bebas Neue or Poppins), position (top, middle or bottom), uppercase, size from 0.5 to 2, and text_color, highlight_color and outline_color as #RRGGBB, #RGB or a color name.

What comes back

A finished job has 5 files. video.mp4 is the video with the captions burned in. transcript.json holds the text, the detected language, the duration and every word with its start and end in seconds. captions.srt and captions.vtt hold sentence cues; the sample has 3. captions.ass is the styled word-by-word file the burn-in uses.

The links stay up for 7 days. Add ?wait= to the request URL to get the finished job in the response, up to 300 seconds, or pass webhook_url to get a signed POST when the job finishes.

What it costs

What the sample cost

Output length
10.5 s
Rate
$0.03 per output minute
Transcription
Included
This clip
$0.0052

Billed per second. A 60-second clip costs $0.03, and 1,000 clips of 30 seconds cost $15.

Limits

What one request takes

Video length
Up to 15 minutes
Input file
Up to 500 MB
Job time
Up to 10 minutes
Spelling hint
Up to 500 characters
Your own words
Up to 50,000
Jobs at once
3 per key
Jobs per day
500 per key per UTC day
Output links
7 days

How it works

From request to file

  1. Get an API key

    Create an account and copy your key. It starts with rsdk_live_ and goes in the Authorization: Bearer header. New accounts get $0.50 of free credit, enough for 10 clips of 10 seconds.

  2. Send the video URL

    POST the JSON body to https://api.reelsdk.com/v1/captions with video_url and, if you need them, style, language, prompt, words, quality and webhook_url.

  3. Wait or take a webhook

    Add ?wait= to hold the request open until the job is done, up to 300 seconds, or pass webhook_url and get a signed POST when it finishes, delivered in up to 6 attempts over about 7 hours.

  4. Keep the files

    Download video.mp4 for TikTok, plus the SRT, VTT, ASS and transcript if you need them. The links stay up for 7 days.

FAQ

Questions, answered

What does a 60-second TikTok cost?

$0.03. Captions cost $0.03 per output minute, billed per second, and failed jobs are free. The 10-second sample on this page cost about half a cent, and 1,000 clips of 30 seconds cost $15. New accounts start with $0.50 of free credit.

Can it show one word at a time?

Yes. Send style as an object with max_words set to 1, for example {"preset": "bold", "max_words": 1}. The sample run with that setting shows one word at a time, and its captions.ass has 27 dialogue lines, one per word.

Which languages does it caption?

Whisper detects the spoken language from the audio, or you can set it with a code such as "en" in the language field. The detected language comes back in transcript.json.

How long does a captions job take?

The 10-second sample came back in 10 to 14 seconds depending on the preset, transcription included. A job can run for up to 10 minutes, on video up to 15 minutes long.

Do I also get SRT and VTT files?

Yes. Every job returns captions.srt and captions.vtt with sentence cues (3 for the sample), plus captions.ass with the word-by-word styling and transcript.json with a start and end time for every word.

Ready to build?

Ship video from your app today

Your first $0.50 is free, and one key covers every endpoint.

Moonlit lake with a lit cabin under the mountains