TikTok, Reels & Shorts Video Analyzer: hooks, scenes, text
Pricing
Pay per event
Every scraper hands you the metadata. This one watches the video and tells you why it worked.
Direct media file URLs (ending in .mp4 and similar), analyzed straight away. Social page links (tiktok.com, instagram.com, youtube.com) will NOT work: those platforms block video fetching from cloud servers, so chain a scraper and use the dataset field above. If you paste a page link anyway, the Actor tells you which scraper to run.
The recommended input. Run a TikTok, Instagram or YouTube scraper first, then pick its dataset here: this Actor reads the media URLs the scraper already collected and analyzes each one. Recognised fields include videoUrl, videoUrlNoWaterMark, mediaUrl, webVideoUrl and url.
Hard stop on how many videos this run analyzes, so a large input list cannot spend more than you expect. At $0.04 per analyzed video, 100 videos is a $4.00 ceiling.
How many stills are sampled and sent to the vision model. More frames means a more accurate read of structure and on-screen text, and a higher bill from your model provider. Frames are always dense over the opening seconds and spread across the rest.
Transcribe the text visible in the frames into the onScreenText field. Turning this off trims what the model writes back, which slightly lowers your model bill. It does not change how many frames are read.
The language for the written fields (hookNotes, whyItWorked, visualSystem, and so on). Classification values such as hookStyle stay in English so they can be filtered and counted.
Your own Anthropic API key, from console.anthropic.com/settings/keys. Analysis runs on your account and you pay Anthropic directly, at their rates. The key is used for the run and is never stored or logged. Leave this empty for a free preview: the videos are still downloaded and measured (duration, resolution, scene count, cut cadence) and you are charged nothing per video.
Which API your key belongs to. v1 supports Anthropic only; more providers are planned.
Leave blank for claude-sonnet-5, the balanced default. Set claude-opus-5 for the most capable read, or claude-haiku-4-5 to lower your own inference bill. The id is checked against your account before the run starts, and an invalid one fails immediately with the list of models you can use.
