One 35 s upload: $0.002 on ReelSDK

Best video API for Lovable, Bolt, Replit and Base44 apps in 2026

Converting a 35-second WebM upload to MP4 with a thumbnail costs $0.002 or less on ReelSDK, $0.014 on Creatomate and $0.025 on FFmpeg Micro. Lovable, Bolt and Base44 run your backend in short serverless functions, and the Supabase Edge Functions behind Lovable Cloud allow 2 s of CPU time and 256 MB of memory per request, so the encode belongs in an API call.

One 35-second WebM to MP4 plus a thumbnail
  • ReelSDK, measured$0.002
  • Creatomate, 200,000 credits yearly$0.014
  • FFmpeg Micro Pro$0.025
  • Shotstack Volume$0.029

Lowest published rate per API. Rendobar bills ffmpeg by the compute second but doesn't publish the rate, so it has no bar.

Head to head

5 video APIs for AI-built apps side by side

Prices are for one 35-second 640x480 WebM converted to MP4 with a thumbnail, from each vendor's rates on 7 November 2026. ReelSDK's rate is on the pricing page; setup guides for Lovable, Bolt, Replit and Supabase show the same call.

5 video APIs for AI-built apps side by side
API35 s WebM to MP4Finished file in one callDocs for AI buildersFree tierWebhooks
ReelSDK$0.002 or less (measured)Yes, ?wait= up to 300 sllms.txt, OpenAPI, open-source MCP serverFirst $0.50 freeHMAC-SHA256 signed, 6 attempts
Creatomate$0.014 to $0.073No, poll or webhookllms.txt, hosted MCP server50 credits, previews up to 480 pxSigning not documented
FFmpeg Micro$0.025 to $0.028No, poll the jobllms.txt, OpenAPI, MCP server200 tokens, 100 MB filesNone, jobs are polled
Shotstack$0.029 to $0.12No, poll or webhookllms.txt, MCP server10 free creditsNot signed
RendobarRate not publishedTwo calls: submit, then get_job waitsllms.txt, hosted MCP server$5 signup credit, 1 job at a timeHMAC-SHA256 signed

Creatomate is priced at 25 fps. Ranges run from each vendor's lowest published rate to its entry plan.

Be honest

Which video API to call from an AI-built app

  • ReelSDK

    One call that returns the finished file

    $0.002 or less for the 35-second conversion, back about 6 seconds after the request. ?wait= holds the request up to 300 s; longer jobs report to a signed webhook. Docs ship as llms.txt, OpenAPI and an open-source MCP server.

    Call ReelSDK from Supabase
  • Creatomate

    Template videos from app data

    $0.014 to $0.073 for this job (about 2.7 credits at 640x480 and 25 fps), with llms.txt, a hosted MCP server and 50 free credits. You poll or take a webhook; webhook signing isn't documented.

    ReelSDK vs Creatomate
  • FFmpeg Micro

    ffmpeg commands with a free tier

    $0.025 to $0.028 for 3.5 tokens, and the free 200 tokens cover about 33 input minutes with files up to 100 MB. It ships llms.txt, OpenAPI and an MCP server; you poll each job.

    ReelSDK vs FFmpeg Micro
  • Shotstack

    Timeline edits beyond a convert

    $0.029 to $0.12 for 35 seconds at $0.05 to $0.20 a minute, 10 free credits, llms.txt and an MCP server. You poll or take a webhook; payloads aren't signed.

    ReelSDK vs Shotstack
  • Rendobar

    Signed webhooks, if an unpublished rate is fine

    Bills ffmpeg jobs by the compute second but doesn't publish the rate, so this job can't be priced. The Free plan has a $5 signup credit and runs 1 job at a time; webhooks are HMAC-SHA256 signed.

    Rendobar pricing

Run it

One call from a Supabase Edge Function

The edge function a Lovable or Bolt app would deploy: one ReelSDK call with ?wait= that returns the MP4 and thumbnail URLs, plus the same body as curl. It ran on 7 November 2026; press Try it to run it in the playground.

supabase/functions/convert/index.tsTry it
Deno.serve(async (req) => {
  const { url } = await req.json();

  const response = await fetch("https://api.reelsdk.com/v1/ffmpeg?wait=120", {
    method: "POST",
    headers: {
      Authorization: `Bearer ${Deno.env.get("REELSDK_KEY")}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      inputs: { "recording.webm": url },
      args: [
        "-i", "recording.webm",
        "-c:v", "libx264",
        "-preset", "veryfast",
        "-crf", "23",
        "-pix_fmt", "yuv420p",
        "-c:a", "aac",
        "-b:a", "128k",
        "-movflags", "+faststart",
        "recording.mp4",
        "-ss", "3",
        "-frames:v", "1",
        "-q:v", "3",
        "thumb.jpg"
      ],
      outputs: ["recording.mp4", "thumb.jpg"],
    }),
  });

  const job = await response.json();
  return Response.json(job, { status: response.status });
});
convert.shTry it
curl "https://api.reelsdk.com/v1/ffmpeg?wait=120" \
  -H "Authorization: Bearer $REELSDK_KEY" \
  --json '{
    "inputs": {
      "recording.webm": "https://api.reelsdk.com/f/in/cipv51ab1dbfgykq/recording.webm"
    },
    "args": [
      "-i", "recording.webm", "-c:v", "libx264", "-preset", "veryfast",
      "-crf", "23", "-pix_fmt", "yuv420p", "-c:a", "aac", "-b:a", "128k",
      "-movflags", "+faststart", "recording.mp4", "-ss", "3", "-frames:v",
      "1", "-q:v", "3", "thumb.jpg"
    ],
    "outputs": ["recording.mp4", "thumb.jpg"]
  }'
Kennedy at Rice University, 1962 (public domain), as a 35-second WebM upload, back as an MP4 and thumbnail in about 6 seconds for a fifth of a cent or less.Input: recording.webmrecording.mp4thumb.jpg

Method

How we ranked them

Five criteria, applied in the same order to every API: the price of one conversion job, whether the call fits inside a short backend function, docs an AI builder can read (llms.txt and an MCP server), the free tier, and webhooks with signatures. An API that doesn't publish its rate ranks last.

The job converts a 35-second 640x480 WebM to MP4 plus a thumbnail, at each API's lowest published rate with the entry price next to it: under 6 seconds of compute on ReelSDK at $0.02 per compute minute, about 2.7 credits on Creatomate at 25 fps, 3.5 tokens on FFmpeg Micro and 35 billed seconds on Shotstack.

We checked each request model against where these builders deploy: Supabase Edge Functions behind Lovable Cloud (2 s of CPU, 256 MB, a 150 s or 400 s wall clock), Bolt's server functions, Base44's Deno functions with a 5-minute limit, and Replit Autoscale at 18 compute units per CPU second.

FAQ

Video API questions from AI app builders

Why does ffmpeg fail in a Lovable or Bolt app?

Their backend code runs in edge functions, not on a server you control. The Supabase Edge Functions behind Lovable Cloud allow 2 s of CPU and 256 MB per request, too little to encode video, so the function should call a video API.

Can a Supabase Edge Function wait for a render?

Yes. Waiting on a network call is async I/O, which doesn't count toward the 2 s CPU limit, and the wall clock is 150 s on Free and 400 s on paid plans. ReelSDK's ?wait= holds the request up to 300 s; our 35-second conversion came back in about 6 seconds.

Which video API works from a Base44 backend function?

Any API you call over HTTP, since Base44 backend functions run on Deno for up to 5 minutes. ReelSDK returns the finished file in that call with ?wait=; Creatomate, FFmpeg Micro and Shotstack start a job you poll or get a webhook for, and Rendobar needs a second call that waits.

Do these video APIs have MCP servers?

All five publish an llms.txt and an MCP server. ReelSDK's MCP server is open source and runs with npx; Creatomate and Rendobar host theirs, and Shotstack and FFmpeg Micro run MCP servers too.

What is the cheapest API for converting user uploads?

For our 35-second 640x480 WebM to MP4 plus a thumbnail: ReelSDK at $0.002 or less, billed by the compute second at $0.02 a minute. Creatomate is next at $0.014, then FFmpeg Micro at $0.025.

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