The post's facts and media — contentType, title, caption, author, stats, direct media URLs, plus an inline thumbnail. Use when you want the post itself and nothing interpreted.
The words actually spoken. Where the platform publishes a caption track (TikTok, Douyin, YouTube) it is read as-is — exact rather than inferred, and far cheaper than watching the video; everywhere else the post's own audio is transcribed, which needs speech-to-text configured on the server. Use bef…
Frames sampled evenly across a post's video, returned as **images you can actually look at** — not a description of them. ffmpeg opens the stream directly rather than downloading it, so HLS works and an expired link is re-resolved on the spot. Verified live at 3/3 on TikTok, YouTube, Instagram, Dou…
Top comments plus the themes the platform clusters them into, with which ones the creator pinned or liked. Use when you want to read what people wrote.
The post's transcript, caption and stats — everything but the pictures, which is what makes it the cheap read. Two fetches: `get_social_media` (1) and `get_post_transcript` (1). The sensible default.
Frames sampled across the video, as **images your model can actually look at**, plus the transcript. Two fetches: `get_post_frames` (2) and `get_post_transcript` (1). Use when the visuals are the point — framing, editing, on-screen text.
The same two fetches, asked for a description of what physically happens on screen rather than why it works. Use when you need the events, not the strategy.
The comment section, every comment with a stable id, and the taxonomy to label them with — sentiment, and whether each is praise, a complaint, a bug report, a question, a request, a comparison or spam. The same `get_post_comments` call, at the same price as reading them directly.
Draws the classifications your model produced — every comment with its sentiment and category, filterable and selectable. Makes no requests; it only renders what you pass it.
Draws the analysis your model wrote from `analyze_post`/`analyze_post_fast`/`understand_social_post` — hook strength, script structure, quotable lines, suggested hashtags and more, whichever fields you produced. Makes no requests; it only renders what you pass it.
The first of two to five posts, fetched with its stats, and the comparison left to you. Fetch the rest with `get_social_media` at 1 credit each. Use when performance differs and you need to know why.
Draws the comparison your model wrote from `compare_posts` — each post scored, the winner marked, what differed and the next test worth running. Makes no requests; it only renders what you pass it.
+56 more tools listed on main page