data-tools

An OpenAI Whisper API alternative without the 25 MB limit

OpenAI’s speech-to-text endpoint takes uploads of at most 25 MB, so an hour of audio has to be compressed or cut into chunks first, and you upload every file yourself. This tool takes links (any size), downloads and transcribes them with open-source Whisper, and returns text, timestamps and subtitles. The price per minute is in the same range as OpenAI’s whisper-1; OpenAI’s newer models are cheaper, so the reasons to switch are the workflow ones below, not price.

What you get

Python example

from apify_client import ApifyClient  # pip install apify-client (3.x)

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("fguiraud/audio-video-transcriber").call(run_input={
    "sources": [
        {
            "url": "https://archive.org/download/gettysburg_johng_librivox/gettysburg_address_64kb.mp3"
        }
    ],
    "model": "base",
    "outputs": [
        "text",
        "srt"
    ]
})

for item in client.dataset(run.default_dataset_id).iterate_items():
    print(item["language"], item["durationSeconds"])
    print(item["text"][:300])

No Python? Open the Actor page, paste your input in the form and press Start; results export to JSON, CSV or Excel.

Real output

Fields from a real run on 2026-10-19 (long values shortened):

{
  "source": "https://archive.org/download/gettysburg_johng_librivox/gettysburg_address_64kb.mp3",
  "status": "ok",
  "model": "base",
  "language": "en",
  "durationSeconds": 157.88,
  "billedMinutes": 3,
  "text": "This is a Librevox recording. All Librevox recordings are in the public domain. For more information or to volunteer, please visit Librevox.org.\n\nThe Gettysburg Address by Abraham Lincoln, November 19th, 1863.\n\nFour scor...",
  "srt": "1\n00:00:00,050 --> 00:00:01,910\nThis is a Librevox recording.\n\n2\n00:00:02,410 --> 00:00:04,850\nAll Librevox recordings are in the public domain.\n\n3\n00:00:05,490 --> 00:00:09,970\nFor more information or to volunteer, plea..."
}

Pricing

$0.006 per audio minute, rounded up per file. Speaker labels add $0.003 per minute.

FAQ

Is accuracy the same as OpenAI?

It is the same open-source Whisper family (tiny, base, small); small is the most accurate option here. OpenAI’s hosted models may do better on hard audio.

When should I stay with OpenAI?

For short files you already have locally, the cheapest OpenAI model costs less per minute.

Can an AI agent use it?

Yes. It is an MCP server at https://mcp.apify.com/?tools=fguiraud/audio-video-transcriber (listed in the official MCP registry), and there is an agent skill for Claude Code, Codex and Cursor. Field-name guesses such as url or urls are accepted.

Ready-made examples