FAQ
Short answers to the questions we get most. Something missing? The quickstart and pricing pages cover the rest.
Which platforms are supported?
YouTube (including Shorts), TikTok, Instagram Reels and many more video platforms — plus photo posts and carousels (an Instagram photo or a TikTok slideshow extracts from its images and caption), and ordinary web pages, which take an article path: we read the page text and images instead of downloading a video. Any publicly accessible URL is worth trying. The probe (POST /v1/probe, or the home page in the web app) tells you immediately whether we can work with it.
Why did my run fail?
The usual causes: the video is private or was removed, the URL points to something we don't support (a channel page, a playlist), or a transient error happened while fetching. Whatever the reason, failed runs are always refunded in full, automatically, so retrying costs you nothing extra.
What if the video doesn't match my schema?
That's a normal, successful outcome, not a failure. If you run, say, the recipe schema on a video that isn't a recipe, the AI won't invent one: the result comes back with outcome: "no_match" (or "uncertain" when the inputs were too thin to be sure), extraction: null, and a one-sentence verdictReason explaining why. In the web app you'll see a "No match" card instead of extracted data. Because real work was done to reach that verdict, the run is billed like any other and isn't refunded. Tip: adding audio or video to your inputs gives the model more to judge and cuts down on uncertain.
Why is my transcript not in English?
Transcripts are built from the video's captions in the video's original language. AI extraction understands that language just fine, so if you want the structured output in English (or any other language), say so in your schema's field descriptions, e.g. "summary of the recipe, in English".
Why was I charged less than quoted?
When a platform doesn't report the video's duration up front, we quote a worst case of 5 minutes. During the run we measure the real duration and refund the difference automatically. Artifacts that turn out unavailable (no captions, comments off) are refunded the same way. That refund is what you see in your credit history. You're never charged more than the quote, only less — that holds even if the content turns out longer than the worst case.
What's a cached result?
If you ran the exact same extraction (same video, same inputs, same schema) within the last 30 days, we serve the stored result instead of re-running everything. Cached results are instant and free. If you want a fresh run anyway (say, to pick up new comments), set force: true on the request (or tick the force re-run option in the web app).
How do I get more credits?
Head to Billing. Your balance, full credit history and top-up options live there.
Are there limits?
Yes, three: videos longer than 60 minutes are rejected at submit, before anything is charged, at most 2 runs at the same time per account, and fair-use rate limits on how fast you can submit new runs. The last two exist to keep extractions fast for everyone; normal usage never hits them.
How do I get help, or send feedback?
Three ways, all of which reach us:
- Live chat — the bubble in the bottom-right corner of every page. It answers straight away and passes anything it can't solve to a human.
- Contact form — pick a topic, write a message, get a reply by email. Signed in, it already knows who you are; signed out, it just asks for your address.
- Email — [email protected], replied to within one business day.
Bug reports, billing questions, a URL that won't extract, or an idea for what Post Reef should do next are all equally welcome. If a specific run went wrong, send us its link — the "Report this" button on a failed extraction attaches the run id for you.