Send a written script to ReelSDK's POST /v1/shorts and get a narrated vertical video back: an AI image per scene, captions and music, at $0.30 per short up to 60 s or $0.45 up to 90 s.

Turn a written script into a narrated vertical video

ReelSDK keeps your script as written, reads it in one of 40 voices, draws a matching image for each scene and returns a 1080x1920 video in one to two minutes.

Use casesUpdated

A 90-word script about Roman concrete in 8 scenes, ready in under a minute.postercaptions.srt

The request

POST /v1/shorts

Send a written script and get a narrated vertical video back: your words read aloud over AI images, with captions and music.

terminalTry it
curl "https://api.reelsdk.com/v1/shorts?wait=300" \
  -H "Authorization: Bearer $REELSDK_KEY" \
  --json '{
    "script": "Roman concrete has outlasted empires. Harbor walls poured two thousand years ago still stand in the sea, while modern concrete can crumble within decades. For a long time, nobody knew why. Then researchers looked closer at tiny white lumps of lime in the old mix. They were not mistakes. When water seeps into a crack, those lumps dissolve and fill it with fresh crystals. The concrete heals itself. Today, engineers are testing the same trick in new buildings, so the next bridge might repair its own cracks, the Roman way.",
    "aspect": "9:16",
    "style": "cinematic",
    "captions": "bold"
  }'
request.mjsTry it
const res = await fetch("https://api.reelsdk.com/v1/shorts?wait=300", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.REELSDK_KEY}`, "Content-Type": "application/json" },
  body: JSON.stringify({
    "script": "Roman concrete has outlasted empires. Harbor walls poured two thousand years ago still stand in the sea, while modern concrete can crumble within decades. For a long time, nobody knew why. Then researchers looked closer at tiny white lumps of lime in the old mix. They were not mistakes. When water seeps into a crack, those lumps dissolve and fill it with fresh crystals. The concrete heals itself. Today, engineers are testing the same trick in new buildings, so the next bridge might repair its own cracks, the Roman way.",
    "aspect": "9:16",
    "style": "cinematic",
    "captions": "bold"
  }),
});
const { job } = await res.json();
response.json
{
  "job": {
    "id": "job_thtrbf9guyzr8iq780te",
    "op": "shorts",
    "status": "succeeded",
    "stage": null,
    "created_at": "2026-11-07T01:31:37.319Z",
    "finished_at": "2026-11-07T01:32:24.926Z",
    "outputs": {
      "video": "https://api.reelsdk.com/f/out/job_thtrbf9guyzr8iq780te/video.mp4?s=...",
      "transcript": "https://api.reelsdk.com/f/out/job_thtrbf9guyzr8iq780te/transcript.json?s=...",
      "srt": "https://api.reelsdk.com/f/out/job_thtrbf9guyzr8iq780te/captions.srt?s=...",
      "vtt": "https://api.reelsdk.com/f/out/job_thtrbf9guyzr8iq780te/captions.vtt?s=...",
      "poster": "https://api.reelsdk.com/f/in/job_thtrbf9guyzr8iq780te/scene-01.jpg?s=..."
    },
    "result": {
      "title": "The Secret of Roman Concrete",
      "post_caption": "The ancient Romans had a secret for building things that last forever. Discover the magic of self-healing concrete! 🏛️🌊 #history #engineering #ancientrome #sciencefacts #innovation",
      "scenes": [
        {
          "narration": "Roman concrete has outlasted empires.",
          "image": "https://api.reelsdk.com/f/in/job_thtrbf9guyzr8iq780te/scene-01.jpg?s=..."
        },
        {
          "narration": "Harbor walls poured two thousand years ago still stand in the sea, while modern concrete can crumble within decades.",
          "image": "https://api.reelsdk.com/f/in/job_thtrbf9guyzr8iq780te/scene-02.jpg?s=..."
        }
      ]
    },
    "duration_s": 36.033,
    "error": null,
    "usage": {
      "output_seconds": 36.033,
      "audio_seconds": 35.424,
      "tts_characters": 517,
      "images": 8
    },
    "poll_url": "https://api.reelsdk.com/v1/jobs/job_thtrbf9guyzr8iq780te"
  }
}

What the API does with your script

The API reads your script word for word for the narration. It splits your text at sentence breaks into scenes, targeting about one scene per 4 s of narration and between 4 and 24 scenes total. From there, a single model call writes the image prompts for those scenes, plus the title and the post caption; your text stays as written.

Each scene renders as one AI image in your chosen style with a slow Ken Burns move, which makes the output a narrated image sequence rather than filmed footage. You can pick from 8 styles, such as a cinematic film still with an anamorphic 35mm lens and natural film grain, or a dark documentary still with low-key photography and a single harsh light source. In this request, the 90-word script about Roman concrete became 8 scenes.

At a glance

The 8 image styles

Each scene's image prompt starts with the style's words.

The 8 image styles
StyleWhat every image prompt asks for
cinematic, default, this runCinematic film still, anamorphic 35mm lens, natural film grain
photorealUltra-realistic photograph, full-frame camera, 50mm lens
animeJapanese 2D anime key visual, cel-shaded animation frame, clean thin line art
3d-cartoonStylized 3D animated film render, appealing cartoon proportions, soft subsurface shading
watercolorFilm still from a hand-painted watercolor animated feature, soft translucent washes
comicComic book style digital illustration, bold ink outlines, halftone dot shading, cel coloring
dark-documentaryDark documentary still, low-key photography, single harsh light source
oil-paintingClassical oil painting, old master technique, visible impasto brushstrokes

Sources: ReelSDK shorts docs, ReelSDK OpenAPI spec, checked November 2026.

Length and price

A short costs $0.30 when it runs up to 60 s and $0.45 up to 90 s, so the $0.50 signup credit covers 1 short like this one. You can send a script up to 3,000 characters, and if the text runs over 90 s at the selected voice's pace, the API rejects the job before anything renders.

The default voice is luna, at about 2.8 words a second, so these 90 words made a 36-second video. You have 40 English voices and 10 Spanish voices to pick from, and fr, ja, zh and ko speak too. If you only have a subject in mind, send topic with up to 500 characters and the API writes the script. See pricing.

What comes back

When the job finishes, it returns video.mp4, a 9:16 video at 1080x1920, plus a poster image, captions.srt, captions.vtt and transcript.json. The response payload also includes a generated title, which came back here as The Secret of Roman Concrete, along with a post caption with hashtags.

You also get the breakdown of every scene with its exact slice of narration and its generated image URL. All the available request fields, including the 8 styles and 8 music moods, are documented in the shorts docs. If you want to pass in your own photos instead of generating AI images, use a narrated slideshow.

What it costs

What this short cost

Short up to 60 s, this one 36 s
$0.30
Total
$0.30

One flat price covers the voice, the images, the captions and the music. A failed job costs nothing.

Limits

What one request takes

Script
Up to 3,000 characters
Length
Up to 90 s at the voice's pace
Scenes
4 to 24, one per 4 s of narration
Concurrent jobs
3 per key
Jobs per day
500 per key
Output retention
7 days

How it works

From request to file

  1. Write the script

    Pass plain text up to 3,000 characters that reads in under 90 s at your voice's pace.

  2. Pick a style, voice and captions

    Choose one of the 8 styles, a voice like luna and a caption preset.

  3. Send POST /v1/shorts

    Send the request to POST /v1/shorts and add ?wait=300 to get the finished job back on the same connection.

  4. Download the files

    Grab video.mp4, the poster image and the caption files from the response links before they expire in 7 days.

FAQ

Questions, answered

What does it cost?

A short is $0.30 up to 60 s and $0.45 up to 90 s, flat, which includes the voice, images, captions and music. The $0.50 signup credit covers 1 short like this one.

Is it real video footage?

No. Each scene is an AI image with a slow camera move, cut to your narration with captions and music.

Can I pick the voice?

Yes. English has 40 voices with luna as the default, Spanish has 10, and French, Japanese, Chinese and Korean speak too.

How long does a render take?

Usually one to two minutes; this 36-second video took under a minute from request to finished file. Add ?wait=300 to get it back on the same request.

Can I send a topic instead of a script?

Yes. The topic field takes up to 500 characters and the API writes the script for you.

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