Logo

Video Processing

Task progress is available through WebSocket. See Task Progress Query.

Gitruck Cloud video processing APIs provide AI subtitles, rule-based shot segmentation, multimodal shot segmentation, motion-highlight analysis, black border removal, canvas resize, frame interpolation, super resolution, video stabilization, watermark removal, Vaporwave filter effects, project file export, timeline rendering, smart oral cut, video translation dubbing, long-to-short highlights, Character Cut, and B-roll smart arrangement.

General Notes

All video processing APIs use an asynchronous task model:

  1. Upload file: Call the File Upload API to get a file_id.
  2. Create task: Call the corresponding processing API (POST) with the file_id to create a task and receive a task_id.
  3. Poll for result: Call the query API (GET) to poll the task status. Results are available when status becomes completed.

💡 Tip: Video processing tasks typically take longer. Poll every 5–10 seconds.

API List

APIPathDescription
AI Subtitle/task/video_ai_subtitleAuto-generate subtitles with translation and style options
Video Segment/task/video_segmentDetect shot boundaries from visual changes and optionally export clips
AI Video Segment/task/video_ai_segmentSplit video shots with multimodal semantic understanding and structured grouping
Video Motion Cut/task/video_motion_cutDetect highlight intervals with obvious motion or camera movement and return structured results only
Remove Black Borders/task/video_blackborder_removeAuto-detect and remove black borders from video
Video Matting/task/video_mattingGreen-screen-free human video matting to a transparent-background video (WebM/VP9 alpha)
Canvas Resize/task/video_canvas_adaptResize video to a target canvas size with optional clipping
Smart Split-Screen Layout/task/video_split_screenCompose 2-16 video clips into one split-screen video with shared layout templates
Frame Interpolation/task/video_interpolateAI frame interpolation to improve video smoothness
Remove Watermark / Remove Subtitles/task/video_purifyRemove watermarks, logos, or subtitles from video
Video Stabilization/task/video_stabilizerAnti-shake processing for smoother footage
Video Super Resolution/task/video_upscaleAI upscaling to improve video resolution and clarity
Vaporwave Filter/task/video_vaporwaveApply a Vaporwave-style color filter to video
Project File Export/task/video_project_structConvert a timeline structure into Premiere/FCPX/OTIO/JianYing/CapCut project files (no upload required)
Timeline Render/task/video_timeline_renderRender a timeline structure directly into a final mp4 (shares the timeline model with Project File Export)
HTML Animation Render (Layered)/task/html_animate_renderComposite a long base track + transparent HTML animation particle overlays (not covering the subject) into a final mp4
HTML Animation Render (Simple)/task/html_render_simpleComposite a video base track + custom HTML animation particles in one pass (full-screen opaque cut-ins / short clips / pure HTML animation)
Smart Oral Cut/task/video_oral_cutOne-click smart editing for talking-head footage: removes retakes and misreads, keeps the last fluent take, shapes pauses by punctuation
Video Translation Dubbing/task/video_translate_dubTranslates the speech in a video or audio file and re-dubs it with the chosen voice, at one consistent speaking rate with subtitles timed to the dub (replaces audio only, picture unchanged)
Long-to-Short Highlights/task/video_long2shortTurn long content into multiple distinct-topic highlight clips: semantic selection + jump cut + word-level cuts (rough cut)
Long-to-Short Highlights (Pro)/task/video_long2short_proRough-cut core plus blurred-background fit / restrained camera moves / pitch-preserving speed / smart subtitles — one-click finished clips
Audio-Visual Speaker Detection/task/video_speaker_detectDetect visible speakers and fuse ASR, visual tracks, and sentence attribution
Character CutAvailable via partnershipFinds a selected person's appearances across a full video while retaining core shots and useful context (partner-only; no API parameters on that page)
Video Material Search/task/video_clip_searchSearch editable video clips by description/keywords, returning in-clip time ranges (synchronous, not an async task)
B-roll Smart ArrangementAvailable via partnershipArranges your own footage library into B-roll candidate tracks along the script beats: which clip, where, how long; never repeats to fill gaps, footage never leaves your machine (partner-only; no API parameters on that page)
External B-roll SearchAvailable via partnershipSearches public video platforms for B-roll candidates matched to script beats, with low-resolution preview and manual refinement before processing the final selected segments (partner-only; no API parameters on that page)

Supported File Formats

Video processing APIs currently support mp4, avi, mpg, mov, flv, mxf, mpeg, ogg, 3gp, wmv, h264, m4v, and ts. mkv and webm are not currently supported.

Input Video Specifications

Every video processing API checks the input video before the task is created. If a requirement is not met, the request returns an error code; no task is created and nothing is charged:

  • Duration: no longer than 2 hours (otherwise 6019)
  • Resolution: width and height both 4000 px or less (otherwise 6022). Common 3840×2160 video is accepted; 4096×2160 exceeds the limit
  • Frame rate: 120 fps or less (otherwise 6023)

Some APIs have stricter limits; see the notes below and each API's documentation. For the complete limits and error codes, see Media File Limits.

Notes

  • Frame Interpolation: The current implementation validates that the output side length does not exceed 4000 px.
  • Video Segment: Structured shot ranges are returned by default. Set only_struct=false to also export clip files.
  • AI Video Segment: Supports scene, shot_type, narrative, and subject modes, and automatically chunks long videos before merging results.
  • Video Motion Cut: The endpoint is currently fixed to the OpenCVMotionCut engine. No mode selection is exposed, and the API always returns structured results only without exporting clips.
  • Remove Watermark: Supports two paths: one-shot processing when the target is already known, or detect, review, then process when the user needs to choose and edit candidates. It supports ffmpeg and raft; raft uses neural inpainting and supports videos up to 20 minutes.
  • Video Super Resolution: The current implementation only supports videos up to 1 minute, and the upscaled output side length must stay within 4000 px.
  • Video Matting: input video must be within 10 minutes; longer videos are rejected (error code 6019).

Next Steps