This page is available in English.
Pay for what you use,
not for seats.
Trio-Spark is metered on input: outputs are a choice and its probabilities, so there are no generated tokens to pay for. Trio-Lumen is metered on minutes of video labeled. One prepaid balance covers both.
$0.042
per million billed input tokens for text and visual input · output free
Included with every request
Text, image and video. One rate for structured text, one image or camera frame, or a 2–4-frame sampled window.
Full distribution. Every response returns the selected action and the probability of each choice.
Always the current model. Each Trio-Spark release ships to the same API. You get improvements without changing code.
Playground and API keys. Test your own states in the browser, then create a key when you are ready.
$24
per hour of video · billed per minute
Included with every video
Activity and per-hand actions. Timestamped activity segments plus left- and right-hand action and grasp labels for every video.
Quote before you pay. Every batch is quoted first. Nothing is charged until the quote is confirmed, by you or automatically within the budget you set, and never more than the quote.
Only verified results. Videos that are skipped, fail or are held for review are not charged. An identical file your account was already charged for, processed with the same version, is free for 90 days from that charge.
Hosted results or your bucket. Download verified JSON from us, or, once your S3-compatible bucket is set up with us, have results written there.
Running at scale?
For sustained volume, a private deployment in your cloud or on your hardware, or a model situated in your own environment, we scope pricing with you.