Recharm

By Recharm

Overview

Recharm helps marketing and creative teams plan video ads using footage they already own. Give it a creative brief, script, or ad idea and it breaks it into scenes, searches your Recharm clip library for clips that fit each scene, reviews the candidates visually, and returns a shareable shot-by-shot brief with clip picks. It can also find gaps in your footage library and recommend which creator should front a concept. Built for marketers, creative strategists, and video editors producing paid-social and UGC ads. Requires a Recharm account with an existing clip library; it only searches footage in your own library.

Tools

get_brand_info

ChatGPT

Returns a brand's profile: its display name, description, vertical (industry), and website. Unset fields are omitted. Use list_brands to find the brand slug to pass as brandName.

get_clip_poster_image

ChatGPT

Returns the poster image (JPEG bytes) for a clip -- a single still frame. Pass asset: { type: 'Clip', symbol: <clipSymbol> } using the clipSymbol from a search_clips_visually hit. To actually examine the content of a clip, prefer get_clip_sprite_image, which returns a grid of frames sampled across the whole clip.

get_clip_sprite_image

ChatGPT

Returns a sprite image (JPEG bytes) for a clip -- a grid of evenly-sampled frames across the full clip duration. Use this to visually examine what actually happens in the clip (motion, subject changes, scene composition) rather than a single poster still. Pass asset: { type: 'Clip', symbol: <clipSymbol> } using the clipSymbol from a search_clips_visually hit. Not every clip has a sprite generated -- the call will return an error when one isn't available.

list_brands

ChatGPT

Returns the brands the authenticated user can access -- their account's brands plus any brands their account is authorized to access. Each entry has the brand's slug name and a human-readable prettyName. Use whoami for the caller's identity.

list_labels

ChatGPT

Returns all label categories and their labels for a brand, filtered to only labels assigned to at least one clip. Each entry has a category and a labels array of string values.

save_brief

ChatGPT

Renders and uploads the final footage brief to Recharm's public-share storage and returns a sharable URL. Call once, after the brief is final; each call mints a fresh URL. The server derives every clip's media URLs, display name, duration, and aspect ratio from each clipSymbol. A scene with an empty clips array renders a "custom shoot needed" gap. (v1 raw-HTML uploads are no longer supported.)

save_html_file

ChatGPT

Uploads raw HTML content to Recharm's public-share storage and returns a sharable URL. Maximum 20 MB. Use save_brief instead when you want the server to render a structured brief from clip symbols.

search_clips_visually

ChatGPT

Searches a brand's clip library by a natural-language description of visual content. Returns ranked clip hits. Each hit includes: clipSymbol, clipName (<symbol> - <sceneType>), rawVideoPublicId (the source video the clip came from), sceneType, cosineDistance (lower is a closer match), durationMs, previewUrl (short looping preview video), posterUrl (still thumbnail image), downloadUrl (full-resolution final clip video), rawVideoDownloadUrl (full source video the clip was cut from; omitted when unavailable), recharmUrl (direct link to the clip in the Recharm library), and categoriesAndLabels (label categories assigned to the clip, populated for the top hits). Clips that share a rawVideoPublicId are different moments from the same source video and tend to look near-identical -- avoid picking more than one of them for the same scene. Pick clips by clipSymbol -- the brief renderer derives all dimensions from it.

search_transcript

ChatGPT

Searches the spoken transcripts of a brand's raw (source) videos for words or phrases. Returns the most relevant raw videos with timed excerpts of what was said and the clips cut from each video. Use it to find footage of someone saying something. Pick clips by clipSymbol.