Woven Video product guide

Get started on Mac

Woven requires macOS 15 or later. Download the app, sign in to your Woven account, and complete trial or subscription setup. Choose how to access chat models, then create a new workspace or open an existing folder as your workspace. You can continue onboarding without adding a video and import footage later.

Start with a concrete request such as: ‘Turn these clips into a 30-second vertical product demo with captions.’ You can ask for a script or shot list first, bring existing media, or generate assets before assembling the edit.

Projects, files, and versions

A workspace is a folder on your Mac; new workspaces go in Movies > Woven by default. It can hold footage, images, audio, scripts, brand references, notes, and multiple reels. Browse and preview files in the sidebar, and open reel documents in the editor.

Reels are saved as .reel.json documents that refer to their media files. New reels created without an explicit path go in the workspace’s reels folder. Keep the referenced assets with your reel when copying or backing up a project; the reel document does not embed every source file. Separate reel documents can reuse the same assets for different cuts.

Edit by chat and by hand

Ask Woven to assemble or revise the reel open in the editor, then refine it using the timeline and canvas. The agent can read the current editor state, including unsaved edits, and apply validated reel changes. You can continue editing visually after an agent edit. To point the agent at something specific, right-click a clip, a caption, or the captions track and choose Reference in chat; right-clicking inside a multi-selection keeps the selection. Long chats are condensed automatically. Hide or show the agent with the sparkles button or Command-Option-J.

Arrange, trim, split, duplicate, delete, and move clips. Select multiple clips with Command-click, a marquee, or Command-A; use undo and redo to revise changes. Drag media from the sidebar or Finder onto the timeline. Imported files are copied into the reel workspace. Stack picture layers for backgrounds, overlays, and picture-in-picture, and position or resize elements on the canvas. Tab moves focus between the sidebar, chat, and editor; Command-Return sends a message and Command-. stops the agent.

Video, images, framing, and keyframes

Import video (MP4, MOV, M4V, WebM), images (PNG, JPG/JPEG, WebP, GIF; animated GIFs show only their first frame), and audio (MP3, WAV, M4A, AAC, FLAC). Decoding still depends on the file’s actual codec and validity. Combine source video with still images, logos, and overlays. Adjust clip position, size, rotation, timing, and video speed. Choose a 9:16, 16:9, or 1:1 canvas, or set custom canvas dimensions from the Format menu; a blank reel adopts the first video’s dimensions. Picture controls include Fit or Fill, crop and zoom, rounded corners, borders, shadows, opacity, blur, and color (Saturation, Brightness, Contrast, and Warmth, with B&W, Warm, Cool, and Faded presets); images also get In, Out, and Loop animations, while video clips animate with keyframes. The Mask tab shows only part of a video or image with a Line, Mirror, Circle, or Rectangle mask, shaped on the canvas, with Feather, Roundness for rectangle corners, and Invert. Borders sit inside the clip and follow rounded corners and masks. Scale videos and images freely for close-ups (up to 100x, and items can sit partly off-canvas). In the Crop inspector, width, height, and scale are linked.

Use the Framing inspector for zoom effects: inspect the thumbnail, adjust zoom, reset framing, and drag the focal point on the canvas. Review the result during playback before exporting.

Keyframes animate position, size, rotation, and opacity on video, image, and text items, and color, blur, and mask shape on video and image clips. Set a key with the diamond button (Option-K), then drag or edit keys; easings are Smooth, Steady, Speed up, Slow down, and Jump. Keys move with the clip when you move, split, trim, or change its speed, and the agent can keyframe too. Captions, audio, effects, crop, border radius, border width and color, shadow, font size, text-box width, and text color can’t be keyframed.

Native effects and transitions

Woven includes 12 native visual effects that can be added in the editor or requested in chat: Push-in, Pull-out, Punch, Drift, Shake, Focus Zoom, Whip Zoom, Glitch, Flash, Blur Pulse, Whip Pan, and Glass Break. Effects have timing controls so they can land on a reveal, beat, or transition. Overlapping effects do not stack: the active effect with the latest start time takes precedence. A clip’s own blur setting is separate and can combine with Blur Pulse.

Add transitions between adjacent video clips on the same row: fade, slides, wipe, whoosh, flash, zoom blur, pull-in, pull-out, smear, glitch, and TV glitch. The whoosh transition is visual only; add a sound effect for the sound. Transition duration is limited to 0.05 to 2 seconds, and each adjacent clip needs at least two rendered frames. The preview and export use the native renderer. An effect changes the composition; changing a clip’s playback speed is a separate edit.

Captions and text

Generate captions from speech, then edit caption text, timing, and appearance. Apply a style to all captions or customize an individual caption, with a way to revert an individual override. Add separate text overlays for titles, hooks, and callouts.

Caption modes are Highlight, Subtitle, Word by word, Word in place, and Progressive reveal. Controls include font, size, case, alignment, width, color, stroke, and entrance animation. Text and captions can combine Outline, Shadow, Glow, Hollow, and Gradient effects with live previews, text overlays have entrance, exit, and loop animations, and caption animations keep speech timing. Caption entrances can play on the whole caption, each line, or each word, and a Look can blur and dim a caption, a line, or a word. Move the playhead onto a caption to edit its text. Word by word captions show one active word at a time. Progressive reveal accumulates words as they are spoken. Review generated transcription and timings before publishing, especially names and specialist terms. Caption generation shows progress and failure details in the editor. Generating captions again replaces the caption words and resets per-caption styling and positions; the editor asks you to confirm first. Word fixes made by you or the agent are kept, and audio Woven has already transcribed is not charged again. Generate captions before investing in individual caption styling.

Voice, music, and sound effects

Mix source-video audio, narration, music, and sound effects. Trim audio by dragging its edges, split it at the playhead, and adjust or mute its volume; the waveform shows the level you will hear. Ask in chat for fades, looping, ducks that silence a track for a set window, and swells that raise it, for example to make room for narration. Sound effects sit on their own compact SFX row, while voice, music, and other audio share the full-height Audio track.

Browse the built-in library of 35 sound effects, such as whooshes, clicks, and impacts. You can also import your own audio, including narration you record elsewhere: drag it from Finder onto the timeline, where it goes on the Audio track as an audio clip. Woven doesn’t record audio itself. Hosted audio generation covers narration with Eleven v3 text to speech, sound effects with Eleven Sound Effects, and music with Eleven Music. Voice or dialogue can also come from video models with native audio, such as Veo 3.1 or Kling v3, or from ElevenLabs connected on your own account.

Generate images, video, and audio

Use Woven Credits to access the available hosted media models, or choose Connected Providers to use your configured provider integrations. Hosted media models cover images, video, music, text to speech, and sound effects. Generation is requested in chat and runs as a background job; Woven can check its status, and when a generation finishes the agent is notified and continues the work with the output in your workspace.

Supported inputs, durations, resolutions, and prices depend on the selected model. Some models accept reference images or other local assets; relevant inputs are uploaded when required by the service. Check the available model options and current pricing instead of assuming every model supports the same controls. Browse images and videos together in chat and attach any asset as a reference. Built-in guidance for UGC-style and AI-creator videos covers character reference sheets, product-accurate shots, and planning a talking head with B-roll.

Study a reference video

Ask Woven to analyze a local video or a supported public video URL. The reference workflow can download the video, extract frames and audio, transcribe speech with word timings, and attempt song identification with Shazam. It then helps explain the hook, pacing, shot structure, and sound design.

Use the analysis as a starting point for a new script or edit. Download availability depends on the source site and access restrictions; song identification can return no match. The workflow is not limited to short-form source videos, but for longer videos the breakdown covers the first 90 seconds.

Work from a website’s brand

Give Woven a public company website and ask it to extract the brand for a video. The bundled brand workflow uses a browser to inspect the site and collect a logo, colors, font information, and a reusable brand document in your workspace.

Review the extracted assets before using them. Website structure and available brand materials affect what can be recovered. Once saved, the brand reference can inform follow-up requests for intros, ads, and other edits.

Reuse skills and project memory

Woven includes reference-video analysis and brand-extraction skills. It also discovers user skills from supported .woven, .agents, .claude, and .codex skill folders at project and user level. Enable discovered user skills in Settings; bundled skills are available by default.

Use skills for repeatable workflows and project memory for context you want to retain. Woven provides memory read, save, and forget operations, and optional workspace memory sharing with Claude Code when it is available. Review the enabled skills and memory settings for each workspace.

Models and integrations

For chat, choose Woven credits or ChatGPT Codex under Settings > Models > Model access: sign in with ChatGPT to use Codex models on a paid ChatGPT plan, or use Woven-hosted chat on credits, which runs on one default model. Available models and account limits depend on the selected connection. Captions, transcription, and hosted generation are billed from Woven credits even when chat uses ChatGPT. Plans include $10 of credits every month, and these charges draw from your credits, with monthly credits used first.

Connect additional tools through local or remote MCP servers in Settings. Woven supports importing MCP configuration, managing server enablement, and disabling individual tools. Featured connections include fal and ElevenLabs using your own provider credentials. Configuring a server makes its tools available to Woven’s agent; it does not turn the Mac editor into a public web API.

Edit from Claude Code or Codex

Since Woven 0.2.17, Claude Code, Codex (including ChatGPT’s Work mode) and other AI apps that can run commands on your Mac can make videos in Woven. Your AI app can find clips anywhere on your Mac and bring them into a workspace, look at your footage, build and edit reels, check frames, add captions, make images, video and audio, and export. Woven does the editing and rendering, with your Woven account and credits.

To connect an app, open Settings → AI apps in Woven, choose the app, copy the setup prompt, and paste it into that app on your Mac. The same prompts are at https://www.woven.video/setup.md. The prompt adds Woven’s connector, a local MCP server, to the app and opens Woven so you can sign in; it installs Woven with Homebrew only if you agree. When setup finishes, Woven shows it’s connected and tells you how to load its tools: restart the app (Woven offers to do that for desktop apps such as the Claude app or ChatGPT), or start a new session if you use it in a terminal. If Woven isn’t open when the app uses it, Woven starts in the background. Setup needs an app that can run commands on your Mac, so it doesn’t work in chat apps or in cloud and web sessions.

Preview and export

Play and scrub the reel in the native preview, then export an H.264 MP4 through the native renderer with 1080p or 1440p output options; exports requested in chat or from a connected AI app render at 1080p. Choose 24, 25, 30, 50, 60, or 120 fps. Empty reels default to 30 fps. A reel created from a video uses its source frame rate, adjusted to a supported value; importing another clip later does not change the project frame rate. Change it under Frame rate on the Settings page, shown when nothing is selected. Choose the canvas shape for the destination rather than assuming every reel must be vertical.

Exports go to the active workspace’s exports folder with timestamped filenames and never overwrite an existing file; if a save fails, the export is kept in Movies > Woven > Recovered Exports. Track export progress or cancel it in the editor. Only one reel render runs at a time. Review the finished file for picture, captions, and audio before uploading it to your chosen platform. Woven’s editor workflow produces a video file; direct social publishing is not part of this guide.

Local work and online services

Workspace files, reel documents, native preview, and rendering live on your Mac. That does not mean every feature is offline: account access, AI chat, hosted generation, connected providers, web research, and downloads use network services.

Content needed for a model or tool request may be sent to the selected service, including media inputs where required. Settings include file-path and network access controls. Review requested permissions and the relevant provider terms; a local project is not a promise that its contents never leave the device. Automatic error reports, with secrets and home and project folder paths removed, are on by default and can be turned off in Settings > Privacy.

Updates and troubleshooting

Available app updates appear in the top bar, which shows download progress and then Restart to Update when the new version is ready. Woven warns you before restarting during an export.

For missing media, check that the files still exist at the paths referenced by the reel. For a failed generation, inspect its status, provider connection, and balance. For export errors, check the file or step the error names. If access is paused, check the account and billing recovery message; if Woven can’t check your subscription, it offers a retry. Send feedback or contact support with the error and steps that reproduce it.