A poster frame and a contact sheet with one frame per shot come out of one ReelSDK call to POST /v1/ffmpeg: under a fifth of a cent for both from a 30-second 1080p clip, at $0.02 per compute minute.

Video thumbnail API

Yes: on ReelSDK, -ss 13 with -frames:v 1 grabs the poster, and a scene filter with tile=3x2 packs the first frame of each of 6 shots into one 1920x720 sheet.

Use casesUpdated

ReelSDK video API: Six frames from the Artemis II launch clip, one per shot, in a 3 by 2 grid
shots.jpg: the first frame of each of the 6 shots in the 30-second clip. The same job wrote poster.jpg, the full 1920x1080 frame at 13 s.Input: launch.mp4poster.jpg (129 KB)shots.jpg (151 KB)thumb-01.jpg (62 KB)thumb-02.jpg (29 KB)

The request

POST /v1/ffmpeg

Grab a poster frame and a contact sheet with one frame per shot from one ffmpeg call, under a fifth of a cent for both.

terminalTry it
curl "https://api.reelsdk.com/v1/ffmpeg?wait=60" \
  -H "Authorization: Bearer $REELSDK_KEY" \
  --json '{
    "inputs": {
      "launch.mp4": "https://api.reelsdk.com/f/lib/in/use-cases/launch.mp4"
    },
    "args": [
      "-ss", "13", "-i", "launch.mp4", "-i", "launch.mp4", "-map", "0:v",
      "-frames:v", "1", "-q:v", "2", "poster.jpg", "-map", "1:v", "-vf",
      "select='\''eq(n,0)+gt(scene,0.3)'\'',scale=640:-2,tile=3x2", "-frames:v",
      "1", "-q:v", "3", "shots.jpg"
    ],
    "outputs": ["poster.jpg", "shots.jpg"]
  }'
request.mjsTry it
const res = await fetch("https://api.reelsdk.com/v1/ffmpeg?wait=60", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.REELSDK_KEY}`, "Content-Type": "application/json" },
  body: JSON.stringify({
    "inputs": {
      "launch.mp4": "https://api.reelsdk.com/f/lib/in/use-cases/launch.mp4"
    },
    "args": [
      "-ss", "13", "-i", "launch.mp4", "-i", "launch.mp4", "-map", "0:v",
      "-frames:v", "1", "-q:v", "2", "poster.jpg", "-map", "1:v", "-vf",
      "select='eq(n,0)+gt(scene,0.3)',scale=640:-2,tile=3x2", "-frames:v",
      "1", "-q:v", "3", "shots.jpg"
    ],
    "outputs": ["poster.jpg", "shots.jpg"]
  }),
});
const { job } = await res.json();
response.json
{
  "job": {
    "id": "job_vxh4tsmhq1d6a15qgcjw",
    "op": "ffmpeg",
    "status": "succeeded",
    "stage": null,
    "created_at": "2026-11-07T01:20:08.770Z",
    "finished_at": "2026-11-07T01:20:14.907Z",
    "outputs": {
      "poster.jpg": "https://api.reelsdk.com/f/out/job_vxh4tsmhq1d6a15qgcjw/poster.jpg?s=...",
      "shots.jpg": "https://api.reelsdk.com/f/out/job_vxh4tsmhq1d6a15qgcjw/shots.jpg?s=..."
    },
    "result": null,
    "duration_s": null,
    "error": null,
    "usage": {
      "output_seconds": 0,
      "audio_seconds": 0,
      "tts_characters": 0,
      "images": 0
    },
    "poll_url": "https://api.reelsdk.com/v1/jobs/job_vxh4tsmhq1d6a15qgcjw"
  }
}

Pulling a poster frame at a timestamp

Putting -ss 13 before the first -i launch.mp4 makes ffmpeg seek straight to the 13 s mark, and -map 0:v with -frames:v 1 tells it to decode and write just that single video frame. We pass launch.mp4 a second time in args so the shot-detection filter on -map 1:v can still read the file from the start while the poster grab finishes right away.

The -q:v 2 flag keeps JPG quality high on the full-size image. From our 30-second 1080p source, that writes poster.jpg at 1920x1080 and 129 KB.

At a glance

Ways to make video thumbnails

Ways to make video thumbnails
OptionPriceLimitsWhat you run
ffmpeg on your own machineFree; LGPL 2.1 or later, or GPL 2 or later with GPL libraries such as libx264Your CPU, disk and bandwidthffmpeg builds, a job queue, storage
AWS Lambda with an ffmpeg layer$0.0000166667 per GB-second plus $0.20 per 1M requests (x86, US East)900 s per run, 10,240 MB memory and /tmp, 250 MB of code and layers, 6 MB sync payloadThe layer, S3 in and out, retries
FFmpeg Micro1 token per 10 s of input, 500 tokens for $5100 MB input on the free plan, 1,024 MB on Starter and ProAPI calls
ReelSDK$0.02 per compute minute; under a fifth of a cent for this poster and sheet500 MB per file, 1 GB per job, 10 minutes of run timeOne HTTP call

Sources: FFmpeg license and legal, AWS Lambda quotas, AWS Lambda pricing, FFmpeg Micro pricing, checked November 2026.

One frame per shot, or one every 5 s

For the contact sheet on the second input stream, -vf select='eq(n,0)+gt(scene,0.3)',scale=640:-2,tile=3x2 keeps frame 0 and any frame where the scene score goes above 0.3, shrinks each match to 640 px wide (640x360), and packs 6 shots into a 3x2 grid. Adding -frames:v 1 and -q:v 3 writes the grid out to shots.jpg as a single 1920x720 image at 151 KB.

If you want separate files instead of a tiled sheet, a second POST /v1/ffmpeg call with -i launch.mp4 -vf fps=1/5,scale=640:-2 -frames:v 6 -q:v 3 thumb-%02d.jpg pulls a 640x360 frame every 5 s and writes thumb-01.jpg through thumb-06.jpg. That job billed about 2.5 seconds of compute, under a tenth of a cent. When you write an image sequence with a pattern like thumb-%02d.jpg, you list every output file name (thumb-01.jpg to thumb-06.jpg) in outputs, and image sequences need -frames:v set to 20 or fewer.

Thumbnails every 5 sTry it
curl "https://api.reelsdk.com/v1/ffmpeg?wait=60" \
  -H "Authorization: Bearer $REELSDK_KEY" \
  --json '{
    "inputs": {
      "launch.mp4": "https://api.reelsdk.com/f/lib/in/use-cases/launch.mp4"
    },
    "args": [
      "-i", "launch.mp4", "-vf", "fps=1/5,scale=640:-2", "-frames:v", "6",
      "-q:v", "3", "thumb-%02d.jpg"
    ],
    "outputs": [
      "thumb-01.jpg", "thumb-02.jpg", "thumb-03.jpg", "thumb-04.jpg",
      "thumb-05.jpg", "thumb-06.jpg"
    ]
  }'

Cost and limits for thumbnail jobs

The first call, which pulls both poster.jpg and shots.jpg in one go, billed about 3.5 seconds to fetch the input, run ffmpeg and store the files, under a fifth of a cent. Machine start-up isn't billed. Raw ffmpeg on ReelSDK costs $0.02 per compute minute, billed per second with a 1 s minimum, and failed jobs cost $0 (see pricing for the $0.50 of free credit at sign-up).

What it costs

What this job cost

Compute: fetching the input, running ffmpeg and storing the files
3.5 s
Rate
$0.02 per compute minute
Charged
$0.0012

Billed per second with a 1 s minimum. Machine start-up is not billed, and failed jobs cost $0.

Limits

What one request takes

Input files
500 MB each, 1 GB per job
Job run time
Up to 10 minutes
Image sequences
Up to 20 frames per output
Outputs per job
Up to 20
Waiting for a result
60 s by default, up to 300 s
Output retention
7 days

How it works

From request to file

  1. Host the input

    Any public http or https URL on the default port works, as long as it has no username or password in it; ReelSDK downloads it before ffmpeg starts. A local file goes up with POST /v1/uploads?filename=clip.mp4 first: up to 100 MiB per upload, kept 7 days, and larger files go up as a multipart upload.

  2. Send the arguments

    The request reads the clip twice. The first input, seeked with -ss 13, gives the poster: one frame at -q:v 2. The second runs select on the first frame plus every frame with a scene score above 0.3, scales them to 640 px wide and lays 6 out with tile=3x2.

  3. Download the result

    Every file is listed under job.outputs, and the links work for 7 days, so copy what you keep. For long jobs, send ?wait=0 with a webhook_url and fetch GET /v1/jobs/{id} when the webhook arrives.

FAQ

Questions, answered

How do I adjust image quality or switch output formats?

For jpg outputs, set -q:v in args - the examples above use -q:v 2 for the 1920x1080 poster and -q:v 3 for the 640x360 thumbnails. If you want png, webp, avif, gif, or apng instead, change the file extension on the output name in both args and outputs. You can mix formats in the same POST /v1/ffmpeg call as long as the job produces 20 outputs or fewer.

What if I want more or fewer frames, or a different grid size?

Change tile=3x2 in the -vf chain to match the grid layout you want, or adjust the 0.3 threshold in gt(scene,0.3) so select catches more or fewer shot changes. For timed intervals, change fps=1/5 and -frames:v 6 to fit your spacing and count. Just keep -frames:v at 20 or under when writing separate files, and list each expected file name in outputs.

How does the cost compare to FFmpeg Micro and AWS Lambda?

This 30-second video cost under a fifth of a cent for the poster and contact sheet, and a little less for the 6 separate thumbnails. On FFmpeg Micro, jobs cost 1 token per 10 seconds of input (metered per second, with 500 tokens for $5, a 100 MB max input on the free plan, and 1,024 MB on Starter and Pro). On AWS Lambda, you pay $0.0000166667 per GB-second (x86, US East) plus $0.20 per 1M requests.

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