# Editing tools (/editing-tools) Editing tools change a file you already have: an upload or an earlier result. You run them like a model: [check the price first](/how-it-works#price-first), then start the job. Building with the API? [List capabilities](/api/capabilities#list-capabilities) shows each tool's id next to its name. Each tool is a capability: send its id as `capability_id`. Ids: `auto_caption`, `text_overlay`, `trim_video`, `stitch`, `split_scenes`, `merge_layers`, `extract_frame`, `extract_audio`, `change_speed`, `change_voice`, `resize`, `upscale`, `remove_background`, `skin_enhance`, `camera_angle`. ## All tools [#all-tools] | Tool | You send | You get back | Longest input | | ---------------------------------- | ----------------------------------------------------------------------- | -------------------------------------------------- | ---------------- | | Add Captions (`auto_caption`) | `source_video` | Captions burned in | 600 s | | Text Overlay (`text_overlay`) | `source_media` (photo or video), `text` | Your text on the file | Video 600 s | | Trim Video (`trim_video`) | `source_video` | One section | 600 s | | Stitch Videos (`stitch`) | `videos`: 2 to 10 clips | One video, in list order | 120 s in total | | Split Into Scenes (`split_scenes`) | `source_video` | Up to 10 clips, one per scene (min 0.5 s) | 600 s | | Merge Layers (`merge_layers`) | `background` (video or photo), `layers` (1 to 6 photos, clips or audio) | One composited video | Background 120 s | | Extract Frame (`extract_frame`) | `source_video` | One PNG | 600 s | | Extract Audio (`extract_audio`) | `source_video` | One 192 kbps MP3. A video with no sound is refused | 600 s | | Change Speed (`change_speed`) | `source_media` (video or audio) | Faster or slower, pitch kept | 600 s | | Change Voice (`change_voice`) | `source_video`, plus `voice_id` | Same words, new voice | 120 s | | Resize (`resize`) | `source_media` (photo or video) | A new shape | Video 120 s | | Upscale (`upscale`) | `source_media` (photo or video) | A sharper, larger copy | Video 30 s | | Remove Background (`remove_background`) | `source_media` (photo or video) | The subject, no background | No limit | | Skin Enhancer (`skin_enhance`) | `source_image` (a portrait) | A retouched copy | One photo | | Camera Angle (`camera_angle`) | `source_image` | A new camera angle | One photo | * A file slot takes plain asset ids: `"source_video": ["ast_..."]`. * A longer file is refused before the tool runs. Trim it first. Upload sizes: [Limits](/reference/limits). * Change Voice (`change_voice`) takes a voice id (`voc_`) as the top-level `voice_id`, the same as Text to Speech (`tts`). To revoice a recording with no picture, use Speech to Speech (`speech_to_speech`). Voice tools that start from audio, with their fields: [Voice and music](/models#voice-and-music). ## Settings [#settings] Defaults are in bold. The live schema wins: [Read one schema](/api/capabilities#read-one-schema). | Tool | Field | Values | | ------------------------------ | ------------------- | ---------------------------------------------------------------------------------------------------- | | Add Captions (`auto_caption`) | `caption_style` | **`punch`**, `impact`, `boxed`, `word_pop`, `clean`, `marker`, `handwritten`, `terminal` | | | `caption_position` | **`auto`**, `top`, `center`, `bottom` | | Text Overlay (`text_overlay`) | `text` | 1 to 300 characters. Required | | | `position` | **`top`**, `center`, `bottom` | | | `text_align` | `left`, **`center`**, `right` | | | `start`, `duration` | Seconds, video only: start 0 to 600 (**0**), duration 0.1 to 600 (**3**) | | | `font` | **`insta`**, `tiktok` | | | `background` | **`none`**, `dark`, `white` | | | `text_size` | **`default`**, or `custom` with `text_size_px` 8 to 400 | | Trim Video (`trim_video`) | `start`, `length` | Seconds: start 0 to 600 (**0**), length 0.1 to 600 (**5**). Stops at the clip's end | | Split Into Scenes (`split_scenes`) | `mode` | **`auto`** finds the cuts, `manual` makes equal parts | | | `threshold` | 1 to 90 (**17**). Lower finds more scenes. `auto` only | | | `scene_count` | 2 to 10 (**6**): the most in `auto`, the exact number in `manual` | | Extract Frame (`extract_frame`) | `timestamp` | Seconds, 0 to 600 (**0**). Past the end gives the last frame | | Change Speed (`change_speed`) | `speed` | 0.25 to 4 (**1**). Below 1 is slower | | Resize (`resize`) | `aspect_ratio` | **`9:16`** (1080x1920), `4:5` (1080x1350), `1:1` (1080x1080), `16:9` (1920x1080) | | | `mode` | **`fit`** (black bars), `crop` (fills the frame), `smart` (a model extends the picture, photos only) | | Camera Angle (`camera_angle`) | `horizontal` | **`front`**, `left_45`, `right_45`, `left_profile`, `right_profile`, `behind` | | | `vertical` | **`eye_level`**, `low`, `high`, `overhead` | | | `zoom` | **`keep`**, `wide`, `medium`, `close_up` | The other tools take only their files. ## Chain tools [#chain-tools] Put one job's output (its `asset_id`) in the next tool's file slot. Nothing is downloaded in between. Each step is its own job, so [wait for the result](/how-it-works#wait-for-the-result) before you start the next. Split Into Scenes (`split_scenes`) gives back several clips: chain each one on its own. Example: trim a video to 15 seconds, make it vertical, then add captions. **Trim** with Trim Video (`trim_video`). ```json title="Job 1" { "capability_id": "trim_video", "config": { "source_video": ["ast_0193c8f0a1b24e7f9d3c5a6b7e8f0061"], "length": 15 } } ``` **Make it vertical** with Resize (`resize`). The file is job 1's output. ```json title="Job 2" { "capability_id": "resize", "config": { "source_media": ["ast_0193c8f0a1b24e7f9d3c5a6b7e8f0062"], "aspect_ratio": "9:16", "mode": "crop" } } ``` **Add captions** with Add Captions (`auto_caption`). The file is job 2's output. ```json title="Job 3" { "capability_id": "auto_caption", "config": { "source_video": ["ast_0193c8f0a1b24e7f9d3c5a6b7e8f0063"], "caption_style": "boxed" } } ``` ### Exact words with Text Overlay [#exact-words-with-text-overlay] Image and video models still misspell words. For an exact headline, price or brand name, make the picture without the words. Then add them with Text Overlay (`text_overlay`). * One text block per job, typed text only (not a logo). * For a second line, run it again on the file the first job gave back. ```json title="A line at the bottom, on a dark background" { "capability_id": "text_overlay", "config": { "source_media": ["ast_0193c8f0a1b24e7f9d3c5a6b7e8f0064"], "text": "Free shipping this week", "position": "bottom", "background": "dark" } } ```