The Captik API

Your business: balance, what you can spend now (with any credit line), your rates, seats and webhook settings.

{
  "object": "business", "id": "…", "name": "…", "status": "active",
  "balance_credits": 182450.5, "available_credits": 182450.5, "overdraft_credits": 0,
  "low_balance_alert_credits": 1000,
  "rates": { "currency": "INR", "credit_value_inr": 1, "caption_per_minute": 5.5, "export_per_minute": 3.5,
             "export_4k_weight": 2.5, "voiceover_per_minute": 7, "ai_edit_per_minute": 7, "tools_per_minute": 3 },
  "seats": { "used": 3, "limit": 10 },
  "webhook": { "url": "https://…", "events": ["job.succeeded", "…"] }
}

Every charge, refund and top-up in a date range, newest first, with the range's totals on the first page. Plain dates are midnight IST; "to" is not included. format=csv downloads the whole range as a spreadsheet.

Get somewhere to put your video. Up to 100 MB: one PUT to the url returned. Larger files (up to 20 GB) come back as a multipart upload: PUT each part to its url, then call /uploads/complete.

{ "filename": "episode-0412.mp4", "content_type": "video/mp4", "size_bytes": 2147483648 }

// up to 100 MB
{ "upload": { "object": "upload", "key": "uploads/…/….mp4", "method": "PUT", "url": "https://…",
              "headers": { "Content-Type": "video/mp4" }, "expires_at": "…" } }
// larger
{ "upload": { "key": "…", "method": "MULTIPART", "upload_id": "…", "part_size": 67108864,
              "parts": [ { "part_number": 1, "url": "https://…" }, … ], "expires_at": "…", "complete_url": "…" } }

Finish a multipart upload. Send the ETag each part PUT answered with (or leave parts out and we read them from storage).

{ "key": "uploads/…", "upload_id": "…", "parts": [ { "part_number": 1, "etag": "\"…\"" } ] }

Bring a video in and caption it: Captik downloads it (for a video_url), transcribes the speech, times every word and makes caption lines. Charged at the caption rate for the video's length. Send an Idempotency-Key so a retry never makes a second project.

{
  "video_key": "uploads/…",              // from POST /uploads, or:
  "video_url": "https://cdn.example.com/ep.mp4",
  "title": "Episode 412",
  "language": "auto",                     // or a code from GET /languages, e.g. "hi-IN"
  "script": "native",                     // native | roman | english
  "template": "TEMPLATE_ID",              // optional, from GET /templates
  "metadata": { "show": "…", "episode": "412" }
}

202 { "project": { "id": "…", "status": "processing", … }, "job": { "id": "…", "kind": "project.ingest", … } }
// job steps: downloading → extracting_audio → transcribing → aligning → captioning
// job output: { "project_id", "words", "lines", "duration_seconds", "language", "script", "warnings"? }

Your projects, newest first: { object: "list", data, has_more, next_before }.

One project: status (processing | ready | failed), source details, language, script, template, a summary of the edit (including stock_credits), your metadata, and its latest job.

Rename it or change its metadata (keys you send are merged; null removes one).

{ "title": "Episode 412 (final)", "metadata": { "status": "approved" } }

Delete a project. Its open jobs are cancelled.

The captions. SRT and VTT follow the caption lines with real word timings. Add download=1 to get a file. 409 project_not_ready while it is still processing.

{ "object": "captions", "project_id": "…", "language": "hi-IN", "script": "roman", "template": "…",
  "words": [ { "id": "1", "text": "namaste", "start_ms": 820, "end_ms": 1240, "emphasis": false, "low_confidence": false } ],
  "lines": [ { "start_ms": 820, "end_ms": 2310, "text": "namaste doston", "word_ids": ["1", "2"] } ] }

Fix captions: change some words, replace them all, or regroup the lines. Words are in time order, without overlaps, and contain no spaces; lines cover every word in order.

{ "update": [ { "id": "14", "text": "Captik", "emphasis": true } ] }
// or { "words": [ { "text": "…", "start_ms": 0, "end_ms": 400 }, … ], "lines": [ { "word_ids": ["1","2","3"] } ] }
// or { "lines": [ { "word_ids": ["1","2"] }, { "word_ids": ["3","4","5"] } ] }

Switch the script (native, roman, english: free) or translate the captions into another language (charged at the caption rate). Timings are carried over line by line.

{ "script": "roman" }
// or
{ "language": "ta-IN", "from": "original" }

The spoken languages we transcribe (with their scripts) and the languages captions can be translated into.

Every caption template: { id, name, category, needs_cutout, tags }. Templates that draw behind the speaker make a cut-out automatically at export.

The AI Edits looks: { id, name, needs_cutout, bold }.

The music beds you can put under a video or an AI Edit.

The sound-effect library and the automatic sound-design styles.

Voiceover voices, reading styles and the speed range.

Template, colours, sizes, one or two lines, where the captions sit, emphasis, and per-caption changes by line index. GET returns the current style.

{
  "template": "TEMPLATE_ID",
  "colors": { "dominant": "#FFFFFF", "accent": "#FFD400", "text": "#FFFFFF" },
  "font_scale": 1.1, "lines": "double",
  "position": { "x": 50, "y": 78, "width": 80 },        // % of the frame
  "captions": [ { "index": 3, "template": "OTHER_ID", "font": "Bebas Neue" } ],
  "font": "FONT_ID"                                    // your font's id, or a Google or built-in family; null = the template's
}

Add one of your own fonts: a .ttf, .otf, .woff or .woff2 up to 5 MB, sent as a multipart "file" or as a link. Use its id (or its family) as "font" in /style. Free.

// multipart/form-data: file=@BrandSans.woff2, name=Brand Sans
// or
{ "url": "https://example.com/BrandSans.woff2", "name": "Brand Sans" }

{ "id": "…", "object": "font", "name": "Brand Sans", "family": "BrandSans", "format": "woff2", "size_bytes": 48210, "created_at": "…" }

Your fonts, and the Google and built-in families /style takes by name.

Remove one of your fonts. Captions that used it go back to their template's font at the next export.

Remove pauses, filler words ("um", "matlab") and repeated takes. Answers with the cuts made and the edited length.

{ "remove_pauses": { "min_ms": 600 }, "filler_words": true, "repeats": true }

{ "cuts": [[12040, 12890], ...], "removed_ms": 48210, "source_duration_ms": 1500000, "edited_duration_ms": 1451790 }

Automatic sound design, music bed and levels, or mute the original audio.

{ "auto_sound_effects": true, "sound_effects_level": 0.8, "music": "off", "music_level": 0.6, "mute_original": false }

A full AI Edit in one of the looks: Captik watches the video, cuts it, and adds zooms, b-roll, sound and music. Every Customize switch of the editor is here. Send instructions to ask the director for a change, as in the editor. "broll": false takes the stock clips over the talk off; "stock_media": false keeps every stock photo and clip out of the edit, the look's own scenes included (use it for OTT, broadcast or anything that needs rights-cleared pictures; the scenes are drawn without pictures).

{
  "look": "LOOK_ID",
  "customize": { "hook": "3 mistakes new founders make", "zoom": "subtle", "broll": false,
                 "stock_media": false, "sounds": true, "cta": false, "accent": "#FF4D00" },
  "music": "off",
  "graphics": true,
  "instructions": "calmer, fewer zooms"
}

// job output
{ "planned": true, "scenes": 42, "sound_effects": 31, "hook": "...", "music": "off",
  "stock_credits": [ { "kind": "video", "creator": "...", "url": "https://www.pexels.com/video/...",
                       "source": "pexels", "shown": [ { "start_ms": 81200, "end_ms": 83600 } ] } ],
  "notes": [] }

Motion graphics: auto plans them for the talk's moments, or place your own from the library with items. GET lists what is placed.

{ "auto": true }
// or
{ "items": [ { "template": "GRAPHIC_ID", "start_ms": 30000, "duration_ms": 4000, "params": { "title": "..." } } ], "replace": false }

Generate a voiceover from text and place it on the video. Charged at the voiceover rate. GET lists voiceovers.

{ "text": "...", "voice": "warm", "style": "natural", "speed": 1, "start_ms": 0, "volume": 1 }

Cut the speaker out of the background (for templates and looks that draw behind the person). Exports make one automatically when they need it; this lets you prepare it ahead. GET returns its status.

Every tool: its id, where it runs, whether it takes "clips", and its options as JSON Schema.

{ "object": "list", "data": [ { "id": "reduce-noise", "name": "Reduce noise", "group": "voice", "runs": "server",
  "per_clip": true, "description": "…", "options": { "type": "object", "properties": { "strength": { "enum": ["light", "strong"] } } } }, … ] }

The clips of the video, and what each tool has set (null when nothing).

{ "object": "tools", "project_id": "…",
  "clips": [ { "id": "c_main", "own_video": true, "start_ms": 0, "end_ms": 61200 } ],
  "tools": [ { "id": "stabilize", "name": "Stabilize", "runs": "server", "applied": { "clips": [ { "clip_id": "c_main", "level": "recommended" } ] } }, … ] }

Run a tool. The body is its options, plus "clips" if you want only some of the video.

{ "strength": "strong", "clips": { "from_ms": 0, "to_ms": 30000 } }

// job output
{ "object": "tool_result", "tool": "reduce-noise", "project_id": "…",
  "clips": [ { "clip_id": "c_main", "audio_item_id": "a_…", "duration_ms": 30000, "tools": ["reduce_noise"] } ],
  "applied": { … } }

Undo: take off what that tool set. Answers { removed, applied }.

Takes out hiss, hum, fans and room noise, keeping the voice.

{ "strength": "light | strong" }

Studio voice: noise out, rumble cut, clearer, levelled and loud.

{ "intensity": 25 | 50 | 75 | 100 }   // 75 by default

Brings the sound to a set loudness: default (-23 LUFS), tiktok (-12) or youtube (-14).

{ "target": "default | tiktok | youtube" }

Changes the voice. The sound tools stack: each works on the last one's result. Undoing any of them gives the clip its original sound back.

{ "voice": "chipmunk | high | low | giant | robot | radio | megaphone | echo | hall | phone" }

Finds the beats of the music in the sound. It returns them only; nothing on the video changes.

// job output
{ "clips": [ { "clip_id": "c_main", "bpm": 118.5, "beats": [ { "source_ms": 512, "edit_ms": 512 }, … ] } ] }

Makes the colours better with a tone curve that follows the footage.

{ "intensity": 80 }   // 0 to 100

Gives the clips the colours of a reference: another clip, a moment or a stretch of the same video, or a picture.

{ "reference": { "at_ms": 41000 }, "intensity": 80 }
// or { "reference": { "clip_id": "c_…" } }, { "reference": { "from_ms": 0, "to_ms": 5000 } },
//    { "reference": { "image_url": "https://example.com/look.jpg" } }

Smooths grain and low-light noise, keeping edges.

{ "strength": 50 }   // 1 to 100

Finds where the footage cuts to another shot and splits the video into clips there.

{ "mode": "scene | shot" }   // scene: only big changes, at least 2 s apart

Smooths out shaky footage. More stable crops in a little more.

{ "level": "recommended | minimum | most" }

Evens out lights that flicker or brightness that pumps. Use weak if the picture changes fast, strong if it changes slowly.

{ "mode": "timelapse | flash", "level": "weak | rec | strong" }

Retouches faces. Values 0-100; leave them out for the automatic retouch. The faces are found once per stretch of video and reused by the next Retouch or Relight. If no face is found, the job still succeeds and its output says so.

{ "clear": 50, "smooth": 45, "whitening": 15, "blemishes": 0, "dewrinkle": 0, "even": 0, "smile_lines": 0, "dark_circles": 35, "bright_eyes": 25, "teeth": 35 }

Adds light: soft light on the face, coloured backlights, moving creative light. Makes the cut-out first when the project has none, so the light falls on the person.

{ "preset": "soft_face | warm_atmosphere | water_light | morning_sun | neon_bicolor | yellow_purple | teal_backlight | blue_backlight | green_side | red_purple_stroke | mysterious_purple | dark_night_green | rainbow_spot | color_swipe | multicolor_switch | metallic_laser | color_rotation | multicolor_flicker" }

Changes the clips' shape and keeps the person in frame as they move. Replaces Stabilize, Camera tracking and AI movement on those clips. Another ratio on the same clips needs no second look.

{ "ratio": "9:16 | 16:9 | 1:1 | 4:3 | 3:4 | 2.35:1 | 2:1 | 1.85:1 | 5.8-inch", "stability": "flexible | normal | stable", "speed": "fast | normal | slow" }

Follows a face, body, hand or an area you pick, and holds it where it was on the picked frame. Use box to pick which one (fractions of the picture); "custom" needs a box.

{ "target": "face | body | hand | custom", "at_ms": 2000, "box": { "x": 0.4, "y": 0.1, "w": 0.2, "h": 0.25 },
  "fit_to_canvas": false, "lock_size": false, "blur_canvas": false }

Camera moves on the beats and gestures (dynamic, soft, zoom, shake) or follows the speech, pushing in on each sentence and punching in on emphasised words. Another preset on the same clips needs no second look.

{ "preset": "dynamic | soft | zoom | shake | speech", "scale": 50, "rotation": 50, "distance": 50, "strength": 50 }

Measures how the camera moves, so an AI edit's scenes and a room look's words stay in place on handheld footage. Makes the cut-out first when needed. A still or unreliable camera keeps everything where it was planned. An AI edit export measures this itself when it has not been done.

{ "again": false }

// job output
{ "camera_moves": true, "trusted": true, "measured_seconds": 30, "samples": 300, "person_grid": true }

Tracks where the person is in every frame, from the cut-out. AI edits and rebuilt looks use it to place their scenes; a depth look uses it to blur and dim the background behind the person. Exports also do this themselves when it is missing.

{ "aspect": "original | 9:16 | 16:9 | 1:1 | 4:5" }   // the export frame the depth grids are for; default the project's

Render the finished video with everything above. Charged at the export rate for the edited length (4K counts 2.5x). The export.ready webhook carries the link.

{ "aspect": "original | 9:16 | 16:9 | 1:1 | 4:5", "resolution": "720p | 1080p | 4k", "quality": "best | small", "burn_captions": true }

// job output
{ "render_job_id": "...", "url": "https://...", "expires_at": "...", "width": 1920, "height": 1080,
  "duration_seconds": 1451.8, "stock_credits": [ ... ], "notes": [] }

This project's exports, with fresh 24-hour download links.

Your jobs, newest first.

One job: status, step, progress, output or error, credits charged.

Stop a queued or running job (an export it started is cancelled too) and return its credits.

Send a signed webhook.test event to your endpoint, to check your receiver.