Eddie

Renders an edit and returns editorial feedback from a video-understanding model: story, pacing, hook, a cut list with timestamps, audio issues and suggested next actions, as text or as timestamped comments on the timeline. Runs in the background and is polled by id. Uses credits; an unchanged cut returns its cached feedback.

analyze_edit

Adds or clears a subtle motion ramp (Ken Burns push, pan or punch-in) on segments of one edit, shown in the player and applied to renders and exports. Shake and handheld drift appear in the player and MP4 exports only.

animate_segments

Adds or removes labels on source files or multicam groups.

label

Rewrites the project's transcripts into a region's written conventions (British, US, Australian, Indian or New Zealand English; Portuguese from Portugal or Brazil; Swiss German; Canadian French; Spanish from Spain or Latin America) without re-transcribing. It changes spelling, number format and punctuation habit only, never meaning or word timing, and applies the same spelling to transcripts added later. The first call quotes the price; a second call with confirmation applies it. Uses credits per minute of transcript.

apply_regional_spelling

Adds one source to an existing multicam group, measuring its sync offset against the group's reference automatically.

assign_media_to_multicam_group

Returns the progress and result of b-roll subclip runs for a project without starting one; when b-roll analysis is complete the reply also lists the top windows inside each b-roll file (in/out seconds, short description, shot, topic, rating, linked subclip).

get_broll_status

Obscures rectangular regions of the picture on a segment of one edit, to hide a face, a plate, a screen or a document. Each region is a normalized rectangle with its own time window inside the segment and its own style: blur, pixelate or a solid fill. The rectangle is static and follows nothing. Regions are shown in playback and baked into rendered video; timeline and project files for other editing applications cannot carry them and say so.

redact_segments

Creates a downloadable document file from provided text or structured data and returns a download link.

build_document

Stops the remaining files of an in-progress footage import owned by the caller. Completed footage and the partial project are kept, and skipped files can be imported again later. Work already running may finish in the background. Project-level finalization continues normally. Requesting cancellation is free; completed footage retains its normal import charge.

cancel_import

Removes wind, hum, background noise or music from under the voice on segments of one edit, using a voice-isolation model on the source file. The cleaned audio covers the whole source file, is generated once and reused by later edits, plays in the player and is used in rendered video; project files for editing applications keep the original sound. Turning it off is free. Uses credits for each minute of newly cleaned source audio.

clean_up_audio

Applies or clears a color grade on one edit, with optional per-segment overrides. Other edits sharing the same footage are unaffected.

grade_edit

Replaces misheard words or phrases in a source transcript while preserving every word timing. Corrections apply to every matching occurrence unless one occurrence is named, the call is atomic, caption tracks built from the transcript are rebuilt automatically, and all corrections on a source can be reverted.

correct_transcript

Creates a new edit from soundbites selected across one or more sources by segment indices, quoted text or time ranges, with an optional target duration and an optional pacing (tight, natural or breathy) that sets how much of each closed pause the cut keeps. Reports the finished length against the source length.

create_edit_result

Creates a multicam group from cameras of the same take with a reference camera and measured sync offsets, or replaces or removes a user-created group.

create_multicam_group

Creates an edit whose timeline is an ordered sequence of trimmed clips or stills with no transcript spine, for b-roll montages and photo slideshows.

create_video_sequence

Creates an empty project, with no media and no edit, in a workspace the caller belongs to, and returns its id and link. With no workspace named it uses the caller's personal workspace. Free.

create_blank_project

Makes an audio-described version of an edit for blind and low-vision viewers: a new edit beside the original, with each spoken description on the voice-over track and a still-frame pause wherever a description needs more time. The user edits and exports it like any edit, or asks for an MP3 of the described soundtrack instead. The first call prices it and starts nothing; a second call with confirmation charges and starts the work, which finishes minutes later. Uses credits per minute of the edit.

create_audio_description

Sets or clears a crop on segments of one edit, given either as a normalized rectangle or as a close-up placed on the presenter's face detected in the segment's frames, at a chosen zoom. Segments where no presenter is found are left unchanged and reported. Crops are shown in playback and applied to renders and exports.

crop_segments

Cuts the project's existing b-roll sources into subclips selected by the user's written guidance and adds them to the b-roll library. The first call quotes the price; a second call with confirmation starts the run. Uses credits.

generate_broll_subclips

Removes backgrounds from media already imported into an Eddie project. Photo mode saves transparent image assets for motion graphics, free; buildReel also creates or rebuilds a named photo reel. Video mode uses sourceId and optional source start/end seconds: first call quotes credits without starting work; after the user agrees, confirm:true charges and starts a background job. The transparent ProRes download appears in exports when ready, minutes later; it is not placed on the edit. Reuses completed video cutouts free.

cut_out_subjects

Deletes an edit by name, including its captions, b-roll, titles and graphics. Requires confirmation.

delete_edit

Measures the beat grid of music (tempo, beat timestamps and a confidence score) on an edit's soundtrack or on a project audio source, or returns a grid measured earlier. Uses credits only when a fresh analysis runs.

detect_beats

Exports an edit and returns a download link, or delivers it into a connected editing system. Formats include Final Cut Pro, DaVinci Resolve, a native Premiere Pro project, Premiere XML, OpenTimelineIO and a cloud-rendered MP4. Uses credits for timeline exports and renders.

export_edit

Exports one frame of an edit, at a chosen moment, as a JPEG image at the edit's frame size and returns a download link. The image shows the footage only; titles, captions, graphics, colour grade and effects are not drawn. Free.

export_still

Rates the transcript segments of chosen sources against ordered quality tiers and exports a tiered selects timeline for an editing application. Uses credits per hour of footage in scope.

export_selects

Writes the full stored transcript of a source, or of every transcribed source, to a downloadable file (text, Markdown, subtitles, CSV, JSON, PDF or Word) and returns the download links. The server copies the stored words, so long transcripts are complete. Free.

export_transcript

Delivers one 4K video file of a whole edit with every clip upscaled, in a single action, instead of upscaling and then exporting as two steps. The first call prices the whole cut and starts nothing; a second call with confirmation charges and starts the work. A clip this project already holds at that size is reused free. It finishes minutes later, and any clip that did not upscale is named in the result and delivered from the original picture rather than passed off as upscaled. Uses credits per clip that needs upscaling, on that clip's length and size; the render is free.

export_upscaled_4k

Matches the passages a user emphasised in a document they attached to a project — bold, highlight or italic — against that project's transcripts, and returns the matching segments in the order the passages were marked, ready to build an edit from. Each passage is reported as matching one place, matching two places equally well, matching a stretch that also carries unmarked speech, or matching nothing, so the ones that need a decision are named rather than dropped. Free.

find_marked_passages

Searches a project's video sources, interview footage included, for the moments the user describes, such as each goal in a match, by watching the picture with its sound, and returns each moment with its source, start and end seconds, a short description and a confidence. The first call quotes the price; a second call with confirmation starts the search in the background. A finished search can be read again for free. Uses credits per hour of footage searched.

find_moments

Searches a project's transcripts for a phrase, topic or speaker and returns matching segments with their source, timestamps and surrounding context. Can also search sound analysis (music, ambience, effects, silence) on sources that have it.

find_transcript_segments

Returns every label and segment modification recorded on the project's sources and multicam groups.

get_labels

Reads the sound of one source over a window of up to ten minutes: the short-term loudness in LUFS and the level of every 100 ms of audio in dBFS, a noise-floor estimate for the window and the whole file, and the silences at or below a threshold. Times are on the source's own timeline. The first call for a file measures it once in the background and asks to be called again. Changes nothing in the project. Free.

get_audio_levels

Returns the status and result of a project-wide camera re-sync run.

get_multicam_resync_status

Returns the timeline comments on an edit, with author, time and position.

get_comments

Returns the credit balance of the workspace that pays for this project, with the balances of any other pools the caller can see.

get_credit_balance

Returns recent credit transactions for the paying workspace: what was charged or refunded, when, for which operation and which project.

get_credit_history

Returns an edit's timeline: its segments with source timecodes and transcript text, b-roll placements, titles, graphics, captions, voice-over clips, music and per-segment treatments, in a summary or a detailed format.

get_edit

Returns frame images from a source or an edit segment so the footage can be viewed: evenly spaced frames across a clip window, the full frame strip of a clip in pages, or frames for specific transcript segments. Metered by the tokens the images use rather than by a credit price.

get_frames

Returns the camera angles of a multicam edit: the group's cameras with names, roles, sync offsets and the current per-segment assignments.

get_multicam_angles

Reads the metadata of source files: the technical layer captured at ingest (codec, resolution, frame rate, duration, audio tracks, original filename, start timecode, device), any start timecode, frame rate or resolution override beside its ingest value, and the descriptive layer (camera label, scene, shot, take, location, keywords and other fields), with per-field provenance and the current revision number.

get_source_metadata