Every Shorz tool,from the chat you already use

Reads the signed-in account's Shorz credit balance and entitlements — the check an agent should run before any paid generation.

Returns the same numbers as the in-app Usage & Pricing window: what each model costs in credits, plus recent spend.

Step one of sign-in: emails the user a one-time Shorz login code, exactly like the in-app Sign in button.

Step two of sign-in: verifies the emailed code and signs the desktop app in. The header balance updates live.

Reads the credentials the app holds — an optional bring-your-own ElevenLabs key, plus this machine's hardware id.

Checks that a bring-your-own ElevenLabs API key actually works before the app starts using it.

Reads the character balance left on the user's own ElevenLabs account (only relevant with a BYO key).

Creates a new Shorz project folder, ready for panels, assets and a render.

Lists every project on disk with its path — how an agent finds the one to work in.

Reports which project the user currently has open on screen, so the agent edits what they are looking at.

Reads a project's whole settings.json — every panel value the render will use.

Patches settings.json with a targeted deep merge instead of rewriting it, so nothing the agent didn't mention gets lost.

Deletes a project.

Switches the project between 16:9, 1:1 and 9:16 (and optionally its fps) by rewriting the output dimensions.

Returns the base folder Shorz keeps projects in.

Returns the app's resource folder — used to locate bundled files such as the connector itself.

Writes the PromptBar brief — what the video should be — without starting a render.

Lists the main AI models available right now, straight from the live server catalog rather than a hardcoded list.

Picks the main AI model that plans the edit — Claude, GPT or Gemini, whichever the catalog currently offers.

Presses Create Video. Uses the saved brief, or a one-off override that doesn't overwrite what the user typed.

Starts a render from a brief passed in the call and returns immediately — poll for status rather than blocking.

Polls the current and most recent render: still working, finished, or failed.

Stops a render that is in progress.

Edits a finished Auto Edit or Text to Video video from one plain-English instruction — 'cut 0:05–0:08', 'make the subtitles bigger', 'give scene 3 a different picture' — and re-renders a new version with every other AI decision pinned, so nothing you didn't mention changes and existing media isn't re-billed.

Renders a title or subtitle to a PNG so the agent can actually see the type before committing it to a video.

Compiles the Remotion source behind the editor preview.

Renders a Remotion composition to MP4 as a background job — bundling plus rendering can take minutes.

The Settings panel: Auto Zoom (count, strength, speed and the zoom-in/zoom-out effect pools), face tracking, freeze-frame, grayscale, colour and speed.

Subtitle styling: font, colours, background, stroke, position and animation.

The title card: text (emoji allowed), style, placement, background box, strokes and animation.

The B-roll panel: your imported clips, web images, GIFs, AI-generated B-roll (stills or video) and emoji — plus where each sits in frame.

The Audio panel: dubbing, music and sound-effect toggles, reverb, volumes and music fades.

The audio-reactive overlay — spectrum bars, waveform, circular bars, voice blob and six more styles, each with its own colour.

The Overlay panel: turns overlay effects on and picks them by name, validated against what's actually installed.

The Border panel: on/off, width, two colours and an animated border style.

The Text-to-Video panel: source mode, script or audio, transitions, image motion, model and reference images.

The Avatar panel: model, script or audio, voice, avatar image, motion instructions and up to three extra angle shots of the same avatar.

The Podcast panel: script, both voices, both avatars, display style and camera motion.

The Advertisement panel: product and character images, and the total ad length in 10-second storyboard scenes (10–40s).

The Music Video panel: the video and keyframe image models, the cut pace, light-leak flashes on section changes, and which part of the song becomes the video.

Points the Avatar panel at a local photo — the person who will be lip-synced.

Adds another shot of the same avatar from a different angle (max 3), so the presenter isn't locked to one pose.

Saves an avatar image into the project.

Points the Avatar panel at an existing local audio file instead of a typed script.

Saves an audio file as the avatar's voice track.

Removes the avatar's saved audio track.

Sets the interviewer's or interviewee's face for a podcast project.

Sets the ad's product shot or character reference from a local image.

Clears the ad's product or character image.

Adds a typed reference image so a scene keeps its look: a character (same person every scene), an environment (same location), or a global style.

Renames a typed reference or updates the notes attached to it.

Removes a typed character, environment or style reference and its cached character sheet.

Saves reference images to the older flat style list (typed references are preferred).

Removes an image from that older flat style list.

Saves a narration audio file for a Text-to-Video project, instead of generating the voice.

Removes that saved narration audio.

Sets a local song as the Music Video project's song and returns its length, which every cost estimate is based on.

Removes the Music Video project's song (your original file is never touched).

Sets an optional artist or character image, so the lead looks the same in every shot of the music video.

Removes that artist or character image; the render then casts the video from the song and the brief.

Generates images with model, aspect, quality and variation control — and optional face references so the same person comes back every time.

Generates a single scene still through the same stack Text-to-Video exports use — the one to pick when reference-image editing matters.

Turns a local still into a short video clip via Seedance, Kling and friends. Minutes, not seconds.

Direct text-to-video and image-to-video generation through the credit proxy, for agents that want to name the model and poll it themselves.

Speaks a line out loud as a real audio file, so a voice can be auditioned before a whole script is narrated.

Generates an instrumental music track from a text prompt into the audio library — ElevenLabs Music when it must run an exact length (10s to 5 min), or Google Lyria 2 for cheap bed material at a length it picks itself.

Generates one sound effect from a short physical description — a whoosh, a UI click, a riser, an impact — at the length you ask for (0.5–22s), saved as an .mp3 in the audio library. The same engine Auto SoundFX uses inside a render.

Lists the ElevenLabs voices available to the account, including cloned ones on a BYO key.

Transcribes a local video or audio file to text as a background job — long footage takes minutes.

Probes a file for duration, dimensions, fps, audio track and size — the first call before any careful edit.

Cuts a video down to a start and end time.

Crops a video or image to a rectangle in pixels.

Resizes a video or image, keeping the aspect ratio when only one dimension is given.

Fits media inside a target frame and pads it with letterbox or pillarbox bars to exact dimensions.

Rotates a video or image 90, 180 or 270 degrees.

Flips a video or image horizontally, vertically or both.

Speeds a video up or slows it down — 0.5 for half, 2.0 for double.

Plays a video backwards.

Loops a clip a set number of times, or until it reaches a target duration.

Holds on one frame for a few seconds in the middle of a video.

Adds a fade-in at the start and a fade-out at the end.

Turns a still image into a video clip of a given length.

Joins two or more clips into one video, in order.

Strips or mutes the audio track on a video.

Scales a video's audio — 0 mutes, 1 keeps it, 2 doubles it.

Fades the audio in at the start and out at the end.

Pulls the audio track out of a video into a standalone audio file.

Swaps a video's audio track for a separate audio file.

Jump-cuts the dead air out of talking-head footage, keeping the speech either side. Only runs when explicitly asked for.

Cuts one or more time ranges out of a video or audio file and joins what's left, frame-accurately — the manual counterpart to remove_silence.

Makes one small, localised AI edit to an image — a colour or lighting change, an arrow, a highlight — and returns a new file. The original is never touched.

Saves stills from a video as image files so the agent can actually look at the footage — evenly spaced, or at exact timestamps.

Searches My Assets across every tab at once — filter by category, name, date range and count.

Lists the finished videos in the library.

Lists the video files available to a project.

Lists the image files available to a project.

Lists the audio files — music, sound effects and voice tracks.

Lists images downloaded from the web inside the app.

Lists the GIFs downloaded into the library.

Lists every thumbnail Thumbnail Creator has produced.

Lists the fonts installed for titles and subtitles.

Imports files into the open project's library lane — main video, B-roll, sound or music — and syncs the UI live.

Renames an asset in the library.

Deletes a local asset.

Saves a file somewhere permanent, optionally to an exact path with no dialog.

Writes image data out to a file on disk.

Opens the app's native image picker and returns the path the user chose.

Reveals a file in its folder on the desktop.

Reports a local file's size, both human-readable and in bytes.

Checks whether a file is actually there before something tries to use it.

Lists the overlay effects available to the project.

Imports overlay effect files, with or without the file dialog.

Removes a user-imported overlay the same way the app's trash button does — file, list and project selection all stay consistent.

Downloads a video by URL from YouTube, TikTok, Facebook or Instagram — the same validation the Clipping panel uses.

Polls that download until it finishes and hands back the local file path.

Saves a generated video from its CDN URL into the video library.

Saves a generated image into the image library.

Saves a generated thumbnail into the library.

Saves a generated music track into the audio library.

Saves a generated sound effect into the audio library.

Saves a GIPHY GIF or sticker into Downloaded GIFs, ready to import as B-roll. Free.

Saves a public-domain sound effect or music track from an Openverse search into the audio library. Free.

Opens the Thumbnail Creator modal, exactly as clicking it would.

Fills in the modal live — including reference images and a YouTube thumbnail to take cues from.

Generates thumbnails with Nano Banana or GPT Image 2.5 and saves the PNGs under Generated Thumbnails.

Lightweight poll for whether that generation has finished.

Closes the modal.

Opens Animation Studio, same as clicking it in the header.

Lists the chat models Animation Studio can use, from the same live catalog as the PromptBar picker.

Sends a prompt to Animation Studio's chat and returns the generated animation code.

One call: prompt, extract the code, and compile the preview.

The whole thing end to end — prompt (with optional image references), compile, and render an MP4 to a path you name.

Closes Animation Studio.

Lists what Animation Studio has exported.

Removes one export from that list.

Clears the export list.

Lists the user's Canvas graphs, newest first, without opening anything.

Reads one Canvas graph without opening it: every node, its settings and current result, the wires, and what each Export holds.

Creates a Canvas graph, or adds nodes and wires to one, and opens it in Shorz so the user watches it appear. Nothing runs yet.

Runs a graph and returns its exported file. Only out-of-date nodes run; a paid run first returns its itemised estimate and waits for an approved credit limit.

Stops a Canvas run, like the Stop button. Generations already sent finish and are kept; the next run picks up where it stopped.

Chooses which take of a node is used, like the take arrows on the node. Everything downstream re-runs from the picked take next time.

Deletes nodes or wires from a graph, like selecting them and pressing Delete. All-or-nothing; generated files stay on disk.

Copies a graph — every node, wire and take, results included — so you can branch it and only re-run what you change.

Renames a Canvas graph without bringing Canvas on screen.

Deletes a Canvas graph when asked: the file moves to Canvas/.trash and everything it made stays in My Assets.

The one publish tool: post or schedule a video, an image, or text alone to YouTube, TikTok, Instagram, Facebook, X, LinkedIn, Threads and Pinterest.

Returns a hosted OAuth link to connect a TikTok, Instagram, Facebook, X, LinkedIn, Threads or Pinterest account.

Lists the connected accounts and their status, re-synced from the publishing provider.

Disconnects one account, or every account on a platform.

The live per-platform rules before posting — TikTok privacy options and length caps, Pinterest boards, and the format limits everywhere else.

Server-priced credit breakdown before you commit — including the surcharge X applies to captions with a link.

Polls a publish job started in the background until every platform reports back.

The content calendar: every scheduled and published post across all eight platforms, filterable by platform and status.

Checks one scheduled or published post by id — the way to confirm an offline-scheduled post landed.

Moves a scheduled post to a new time, with no new charge.

Cancels a scheduled post and refunds its credits automatically.

Pulls fresh performance numbers for published posts. Free — no credit charge.

Uploads to YouTube with privacy and category set, and can schedule it: the video goes up private and YouTube publishes it at your time.

Uploads a file from disk to a named YouTube channel through the same path the in-app uploader uses, with no upload UI.

Starts the YouTube OAuth flow to connect a channel.

Reports which YouTube channels are connected.

Removes one connected YouTube account.

Signs out of every connected YouTube account.

Searches free Pexels stock video by keyword and hands back direct media URLs.

Searches free Pexels stock photos by keyword.

Browses the most popular Pexels videos right now.

Browses Pexels' editorially curated photo feed.

Fetches one specific Pexels video by id.

Fetches one specific Pexels photo by id.

Searches GIPHY for GIFs — reactions, memes, footage loops — with size filters and a G rating by default. Free.

Searches GIPHY stickers: transparent-background GIFs — arrows, callouts, cut-out characters — that sit on top of the video. Free.

Browses GIPHY's trending GIFs right now, no keyword needed. Free.

Browses GIPHY's trending stickers right now. Free.

Fetches one GIF or sticker by its GIPHY id, with every rendition GIPHY has for it. Free.

Searches public-domain (CC0) sound effects — whooshes, impacts, risers, clicks, ambiences — usable commercially with no attribution. Free, no sign-in.

Searches public-domain (CC0) background music by genre — full tracks or short loops — usable commercially with no attribution. Free, no sign-in.

Live X/Twitter search in plain language — posts, news, people, trends and media, with handle and date filters, answered with x.com citations.

Snapshots the live Shorz window — every interactive element with its name, role, value and screen position, plus where the user currently is.

Scrolls an off-screen control into view and returns its new coordinates.

Clicks a control with a real, visible mouse — and refuses, naming the blocker, if Shorz isn't focused or something covers the target.

Types into whatever has keyboard focus, chunked so controlled inputs don't drop characters.

Presses keys and shortcuts — Escape, Tab, Enter, Ctrl+A, arrow repeats.

Scrolls the mouse wheel over a point, for browsing a long list.

Polls any background job — renders, generations, uploads, transcriptions — until it completes or fails.

Checks whether a newer Shorz version is out, with the release and download links.

The raw internal event log. For renders prefer the status tools; this is for logs and debugging.