Article mode

Every composition has Video, Article, and Moodboard surfaces. Video is the timeline editor; Article is the written guide built from the recording. Use the tabs next to the composition name to switch surfaces. The editor remembers your selected surface when you refresh. Moodboard references can be added before an article exists. Vibe's reference tool adds notes and images; tags do not position, recolor, or connect canvas items. Ask Vibe to position references, change a sticky note's color or text, or connect two references with an arrow. These use dedicated board controls, not tags.

Generate the guide

Switch the header to Article. A new page opens like a Notion page: the title and an empty document you can type into straight away. Above the first line, Write the guide from the video reads the transcript (captions or the Select panel's transcript), writes the steps, and keeps a frame from the video for each one; the shape picker beside it chooses a step-by-step article, a quick start, release notes, or an FAQ. The row goes away once the page has content. The agent on the right can do the same from a sentence.

Edit it like a document

The article is a stack of blocks: headings, paragraphs, numbered steps with a picture, bulleted, numbered, and to-do lists, quotes, callouts, toggles, tables, two-column layouts, code, dividers, images, and YouTube embeds.

• Type / in an empty block for the block menu: Regular text, Heading 1 to 4, Step, Quote, Code, Image, Divider.
• Enter starts a new block; Backspace in an empty block removes it.
• The rail on the left changes the block type, applies bold, italic, underline, strikethrough, and links, aligns the current block left, centre, or right (saved with the block and kept in exports), turns a block into a list, inserts a divider or an image, and undoes or redoes.
• Hover a block for the ⠿ handle: drag it to reorder, or open its menu to turn the block into another kind, duplicate, move, or delete it. The + beside it inserts a block below.
• Select text for the bubble toolbar: bold, italic, underline, strikethrough, code, highlight, link, and turn into.
• Drop or paste an image to upload it in place. Hover a step's picture for Make a GIF, Replace, and remove.
• Guides with three or more headings get an On this page outline on wide screens.
• /Video clip embeds the marked zone (or five seconds from the playhead) as a block that plays that part of the recording inline. On the share page and in exports it shows the frame captured when the block was added.
• Type @ to mention a step, a marker, the playhead time, or a media file. Clicking a mention scrolls to the step or seeks the video.

Edits save with the composition. Undo and redo here are the article's own history, separate from the timeline's.

The theme

The right column's Theme tab (beside Agent while Article or Moodboard is open) decides how the guide reads. Answer what you know (who the guide is for, how it should feel, how long) and press Decide with AI: the copilot proposes a tone, audience, reading level, length, a few formatting rules, and a picture style, reading the transcript, the project's Instructions, its brand vocabulary, and the workspace playbook. It explains the choices in a short paragraph. Change any field by hand; the theme saves with the article and shapes every draft and rewrite.

AI on one block

Select text inside any block and the toolbar that appears has Ask AI: Improve writing, Shorten, Expand, Fix grammar, Simplify, or Change tone with the same five tones as the theme. The copilot rewrites only that block and keeps its id, timing, and picture, so nothing else in the document moves. The block menu on the ⠿ handle has the same Ask AI about this block. Cmd+Z puts the old text back.

The moodboard

Moodboard on the header's Video | Article | Moodboard switch opens a canvas for the guide's references and for making new pictures and clips. Drag the ground to pan and hold Cmd (or Ctrl) while scrolling to zoom. The dock at the top adds what you already have: Upload a file (or drop it on the board), pick from the workspace Library, search Stock photos, or drop a Note with a line of direction. Drag anything where you want it; positions save with the board. Delete removes the selection; Cmd+Z undoes.

The toolbar at the bottom has tldraw's drawing tools and two generators. Generate image and Generate video drop a frame at the centre of the view with a prompt card under it. Write what you want, pick the model (each one says what it is good for and what it costs), the aspect ratio, and for video the length, resolution, and whether it has sound. The card shows the credits the request will use before you press the arrow. A new frame opens with Nano Banana 2 for a picture and Gemini Omni for a clip. Seedance 2.5 covers takes longer than 10 seconds, Kling O3 Character Swap replaces a person in a clip with the one in your photos, and GPT Image 2.5 is there when you want OpenAI's model. While the provider works, the frame shows its place in the queue and a progress sweep; when it finishes, the frame becomes the picture or clip at the same spot. A failed generation shows why, keeps the frame, and refunds the credits; press the arrow again to retry.

Select a picture for Edit (change it from an instruction, with any of the editing models), Animate (turn it into a clip), or More: Upscale, Remove background, Extend the edges, and Describe with AI (a caption, tags, and its colours). Select several pictures for Video from these (a clip guided by all of them; cite them in the prompt as @Image1, @Image2) or Combine into a picture. Select a clip for Edit or Extend. Every generated picture and clip also joins the workspace library and the project's media, so the video editor can use it.

A generation keeps going without you. Refresh the page, close the tab, or open the project on another device and the frame is still there, still counting down, and turns into the picture or clip when the job finishes. Klipeo's server follows the job, not your browser.

Pin to step ties a picture to a step (the step then shows how many references it carries, and the writer agent treats pinned pictures as that step's references). Download saves the selected picture or clip, and Send to Vibe attaches it to the agent chat so you can ask about it or have it placed on the timeline. The bar sits above the selection, or below it when there is no room. The board's colours collect at the top right; Use on the theme sends them to the theme's palette.

The agent

With Article open, the right column is the same Vibe agent you use on the timeline, with the Article writer skill. Its empty state offers the asks that matter here: write the guide from the video, decide how it should read, shorten every step, fix the structure, remove repeated screenshots, or change the tone of the intro. Type your own in the same box.

The agent works on the document you see. It can draft the guide from the transcript (the editor then reads a picture from the video for every step), or write paragraphs from your brief without requiring a transcript. This second path does not request screenshots or infer facts from unseen footage. It can also rewrite the whole guide with one instruction, rewrite one block by its id, decide and save the theme, improve an image prompt, and keep a note or picture on the moodboard. With the Media studio skill it also generates pictures and clips with the board's models (it reads the catalogue first, builds on what is already on the board, and asks before making several clips); its results appear on the board and in the library like yours. It reports what changed, and Undo on the rail brings the previous version back. Pictures stay yours: the agent never invents one. The page's ••• menu (top right of the document) rewrites the whole guide from the video in a chosen shape and offers Add animated GIFs, Remove zooms, and Re-read pictures, which read the video here in the browser.

Share and export

Share → Export article has a format picker with a line on what each is good for: Markdown (docs sites, GitHub, Notion), HTML (a single page with print styles), RTF (Word, Pages, Google Docs), and PDF (the print view; choose Save as PDF). Copy Markdown puts the guide on the clipboard. Pictures link to their uploaded frames. A view-only link created with Show article on renders the guide under the video on the shared page.