Pricing

One price, charged per minute of video. Every job runs the whole analysis. There are no tiers, no options to buy and no subscription. New accounts start with $5 of credit.

$1
per minute of video
billed by the second, rounded up to a cent
Minimum 20¢ per job
$0
a failed job
the hold is released
Never charged
$0
the same video again
served from your account's cache
Never charged twice

A 30-second video costs $0.50. A 10-minute video costs $10. The $5 starting credit covers 5 minutes of video.

Credit

You buy credit on the platform, in $20, $100, and $500 packs or any whole amount from $10 to $5,000. Credit is prepaid and doesn't expire. When the API accepts a job, it holds the estimated price. When the job ends, it charges the actual cost or releases the hold.

Limits

A video can be up to 10 minutes long, 300 MB, and 1280 pixels on its longest side. Split a longer video first; downscale a larger one (the analysis reads nothing finer than that). Each key allows 60 requests per minute by default; you can change that per key on the platform.

What every job includes

The sound

Deepgram times every word and separates the speakers. The result has the words, the sentences they make, the language, and the full text.

The cut

A detector finds every cut and fade from the pixels. Each one is then classified from the picture on either side of it: a hard cut, a jump cut, a punch in or out, a zoom. Shots and scenes follow from those.

The picture

A vision model reads the video four times a second and lists everything added on top of the footage, with its text and its box; the pixels then put each start and end on the exact frame. A second pass finds who the video is about and crops every appearance.

What the result holds, field by field, is in the docs.