See the plan while lofi draws
While lofi builds or edits a board, the chat shows the plan as steps, with each screen ticking off as it lands. When it's done, it folds to one line with the time and stays in the transcript.
While lofi builds or edits a board, the chat shows the plan as steps, with each screen ticking off as it lands. When it's done, it folds to one line with the time and stays in the transcript.
Select a word in a post-it or text element and make just that bold, italic, bigger, or a colour. ⌘B and ⌘I work too.
Screens can layer on each other: a sheet, tray, or toast drawn over the dimmed screen below. lofi draws emoji reaction trays, camera previews, and faces from the doodle library instead of placeholder circles. Arrows that skip a screen route over the row, with captions like "Long press the story photo". Also fixed: library doodles lose their grey tile, saved doodles no longer shrink, and Escape keeps a text edit.
Hit Play on a flow's first frame and walk its arrows one screen at a time, at device size. Click the control an arrow names, tap the chips under the screen, or use the arrow keys. A flow map shows where you are, and variants are tabs along the top.
Example prompts drift under the composer: hover to preview, click to edit before you generate. Below them, four boards from lofi, including a Getting started walkthrough, ready to open and Remix.
Generated flows keep their arrow labels ("Tap Continue"). Arrows meet a screen near its top and stack instead of crossing, with labels on hover. Also fixed: a stalled cache could hang every generation. lofi now falls back within seconds.
Paste a link and say "wireframe this site". lofi reads the page and draws it with the site's own headers, cards, pricing, and FAQ, with the live site parked above the frame to compare.
List your screens and you get exactly those, in that order. Tag one [mobile] and it's drawn at phone width beside the desktop frames. Rules in your brief are checked on every screen.
New SiteHeader and MegaMenu components: brand, links, and button on one line, with real menu columns. Dense variants are actually tighter, not just fuller. Lists show real labels instead of "Item 1". Also fixed: large text shrinking while you edit it, and Option pressed mid-drag now duplicates.
Every element inside a frame as a tree, with an icon per component. Drag rows to reorder or nest, lock from the row, shift-click to add to the selection, and walk it with the arrow keys.
Elements inside a frame always draw, whatever the nested-editing setting. The layers panel no longer hides behind the frame toolbar.
Project cards show a rendered snapshot of the board instead of grey blocks, refreshed after each save.
Type / in the composer for /research, /variants, and /flow. /research is look-only: lofi scans a few well-known products and drops the findings on the canvas as a note card with screenshots. Nothing gets built until you ask.
lofi asks who the product is for, with options written for your product, and recommends a device. Density is inferred. Your first build also writes a project brief that every later generation reads.
Handoff is one zip now: LOFI.md with purpose and acceptance criteria per screen, PNGs of every screen, your attached docs, and ready-made agent files. The dot grid stays crisp at every zoom and frames snap to it. Progress labels read "Sketching screen 3 of 5", not "1 of 1". Groups get one outline, with Group and Ungroup in the right-click menu. Also fixed: the chat going blank mid-generation.
Paste any link onto the canvas and it opens as a mini browser window. Double-click to browse it. Select it and ask lofi about it, and lofi uses the site's structure as reference.
New palette button in the top right. Pick the Clean voice and one of six accent colors. Works on every board you already have, and exports follow.
While lofi works you see the stage, "Planning the flow" then "Sketching screen 2 of 4", with a timer. Failed turns get an inline Try again. Copy any message, or thumbs a reply up or down.
156 hand-drawn doodles, up from 34, and lofi picks ones that match the content. Pasted images and live embeds survive redraws. Canvas edits stay in the transcript after a reload. The top-right controls are one pill: theme, Share, Export.
Anyone with a share link can open your board, no account needed. Everyone sees each other's cursors, and your edits and generations show up live.
Hit Remix on a shared board to get your own private copy, frames and chat included.
The chat, layers, and code panels are resizable now: drag the edge, double-click to reset. Frames rotate from just outside their corners, contents included. And pan, move, resize, and arrow gestures no longer step on each other.
Select a sticky, shape, image, or text and turn it with the new handle below the selection box. Snapping, arrows, and exports all follow the rotated shape.
Cmd+Shift+L locks the selection so it can't be moved, restyled, or deleted. Locked elements stay clickable, so unlocking is one click on the badge. Locking a frame locks everything inside it.
Canvas text takes color and alignment now, and Cmd+I italicizes. The shape tool previews the real silhouette while you drag. Also fixed: a save that could drop your selection mid-edit. And the Generate button got the marketing site's 3D look.
Code is now its own element, FigJam-style: a dark block with line numbers and a language picker, placed straight from the elements picker.
Ask for a landing page and you get one tall page, not a flow of section frames. lofi also asks how many variants you want, and shares the takeaway from its research before it builds.
Rectangles, ellipses, and lines are here: press R and drag, then style fill and stroke from the floating toolbar. Cmd+G groups elements so they move as one. Arrows attach to anything, and double-clicking one adds a label. And pan mode always pans, whatever is under the cursor.
Connect your Mobbin or Landingfolio account under Connections and lofi studies real apps and landing pages while it plans. References shape structure, never branding or copy.
Click any frame and it appears in the composer as a chip. Type what you want changed and send, and lofi edits that frame in place. There's also a new Layers panel in the frame toolbar showing everything inside the frame. The landing page got a restyle to match the hand-drawn look of the canvas, and the composer holds still when you select a frame.
Paste a screenshot with Cmd+V or drag image files onto the canvas. PNG, JPEG, WebP, and animated GIFs all work. And lofi can see them: select an image or drop it into chat and it becomes visual reference while lofi wireframes.
A new switcher in the composer: Default for quick drafts, Thinking for tougher briefs. Thinking mode uses a deeper model and asks a couple of product questions before it builds. Messages now run on AI credits, one per message, allowances unchanged: 30 on free, 500 on Pro.
Common app types like habit trackers, food delivery, and fintech dashboards now come out structured the way real apps are, with realistic content ("12-day streak", not "Item 1"). Also fixed: trackpad pinch zooming the whole page, the first message on a new project failing, and doodle strokes drifting while you draw.
Grab any component inside a frame and drag it to a new spot. The layout reflows live while you drag, so you can see exactly what you'll get before you drop. Nesting things into containers works the same way, and so does pulling them back out.
Exporting to Claude Code (or any coding agent) is now one link. Hit Export, copy the link prompt, and paste it into your agent. It fetches the spec straight from lofi, and if you keep editing, it just fetches again. No more downloading and re-pasting files. The full prompt and markdown download are still there if you prefer them.
Chat now takes PDFs alongside images. Drop in a brief, a spec, or research notes and lofi reads it as context while it wireframes.
Image placeholders don't have to be gray boxes anymore, there's a new Doodles tab in the picker with hand-drawn illustrations. Chat messages can be much longer now (20k characters, with a clear error instead of a silent cut-off). And the emails lofi sends, like confirmations and password resets, finally look like lofi.
When lofi finishes a flow, it now suggests next steps as tappable pills under the reply. Try a variation, flip the flow to desktop, or push a screen further without writing the prompt yourself.
Ask for another take on a flow and lofi builds a full variation next to the original, keeping both. Typing works too: "try a different version" or "make a desktop version" gets you the same result as tapping a pill. Skeleton frames hold the space while the new flow generates.
One tap turns a mobile flow into a desktop one, or the reverse. Same screens, same steps, re-laid-out for the new size.
Some people confirmed their email or reset their password but still couldn't get in, mostly on corporate email setups. Both links now log you straight in.
Every frame now has code behind it. Select a frame, hit the </> button, and copy a Claude-ready prompt for that screen, or grab its code as frame JSON or HTML with Tailwind.
We did a deep pass on output quality. Mobile frames now get real grid layouts, honest padding, and accurate heights. FABs sit bottom right, composers and CTAs stay pinned to the bottom, and there are new components like chat bubbles and richer list items.
One real session, start to finish. Sketch a flow in lofi, shape the screens, then hand it off to Claude Code.
Pick a sticky note, text, frame, or UI element and a semi-transparent preview follows your cursor, so you see exactly where it will drop before you click. Panning and zooming keep working while any tool is active.
The thinking indicator no longer disappears while lofi is still working, returning to a project no longer replays old messages, and long conversations render faster.
You can now attach images to a message in chat. Drop in a screenshot, a hand sketch, or a reference and lofi will use it as context while it wireframes. Up to 5 images per message, each with a full-screen preview so you can check the details.
Today lofi opens its doors. It's the low-fi canvas for wireframing flows — think in frames, describe what you want, and lofi sketches the screens with you before you prompt for code. This is an early alpha, so expect rough edges and fast changes. Here's a quick look at what it can do.