Thirteen extraction tools. Each row states what comes back, what it costs, and the limit that matters — because the limit is usually the thing you needed to know before you started.
All of them return one row per input, including the inputs that produced nothing. A video with no captions, a company that publishes no address, an app with no reviews in a given country: those come back as rows with a reason attached, and are never charged. That is the whole design.
Prices are the list rate. Paid Apify plans get up to 52% off through Store discounts.
Transcripts for individual YouTube videos, in bulk.
Further reading: Reliability benchmark · Why it fails in production · Try one free
Open on ApifyEvery long-form video on a channel, transcribed.
Further reading: Archiving a university channel
Open on ApifyEvery video in a playlist, in playlist order.
Further reading: Turning a course into notes · Reading a conference track
Open on ApifyA channel's Shorts, which most transcript tools skip entirely.
Further reading: Why other tools return nothing
Open on ApifySearch YouTube by keyword, then transcribe what comes back.
Comments and reply threads from any public video.
Further reading: Selling research built on comments
Open on ApifyPublisher-made transcripts with speaker labels and real timestamps.
Further reading: A clip service built on this
Open on ApifyContacts a company publishes on its own site — never guessed.
Further reading: What 400 sites actually publish · How to find contact details · Role vs personal addresses · Selling lead lists
Open on ApifyOpen roles from companies' own boards, across seven ATS platforms.
Apple App Store reviews from any country store, not only the US.
Messages from public Telegram channels, with no bot token.
Posts from any public Bluesky account, with no login.
Posts from public Threads profiles, with no login.
Most extraction tools describe what they return and stop. The result is that you discover the cap, the coverage gap or the missing field after you have built around it. Every limit above is one we measured — see thebenchmarks for the numbers behind two of them.
Nothing here guesses. No invented email addresses, no inferred remote flags, no transcript generated from audio and presented as a published one. Where a source publishes nothing, the row says so.