FLORA now connects directly to Dropbox — import assets and export finished work without leaving your canvas. This week also brings precision zoom controls, a major canvas performance push for large projects, and smarter video generation that fixes inputs instead of rejecting them.

📦 New

• Dropbox Integration - import from and export to Dropbox right from your canvas
  Pull assets in through your library or send finished work out via the export node with a native folder picker. Works across shared workspaces, remembers your last destination, and keeps everything in sync — no more downloading and re-uploading between tools.

• Zoom Controls - precision navigation in the editor and fullscreen view
  Zoom now anchors to your cursor with keyboard and HUD controls, a clear zoom indicator, and dedicated controls in fullscreen image view — so inspecting fine detail in large generations is effortless.

• Reference Images in Inpaint - @mention any image on your canvas inside inpaint prompts
  Guide your edits with visual references instead of describing them — point the model at exactly the style, object, or texture you want painted in.

• FAUNA Understands Video & Audio - connect video or audio nodes and FAUNA gets it
  FAUNA now eagerly analyzes video and audio context on your canvas, so you can ask about a clip's content, pacing, or sound the same way you already do with images.

• Lock Selection - lock nodes in place from the selection toolbar
  Freeze finished sections of your workflow so a stray drag can't disturb layouts you've already perfected.

• Optional Technique Inputs - build more flexible reusable workflows
  Technique creators can now mark inputs as optional, so one technique can serve both quick runs and fully customized ones.

• 8K Video Upscaling - the video upscaler output cap is raised to 8192×8192
  Push your footage to true 8K resolution for final delivery.

🧠 Improved

• Canvas performance - a dedicated optimization push for large projects: smoother panning and zooming, no more hover stutter, faster toolbar interactions, and snappier rendering on canvases with hundreds of nodes.

• Video generation is more forgiving - undersized videos are now automatically upscaled to meet model requirements instead of being rejected, preprocessing is more reliable, and resolution/duration errors are clearer about what to change

• Downloads handle huge images - images over 150MP now download reliably, and "Download original" is always available

• Billing transparency - see a full cost breakdown per generation in your usage page, get a clear notice on your dashboard if there's a billing issue, and overdue invoices recover automatically when you update your payment method

• Project search - finds matches from the first character you type

🔐 API Changes

• Retrieve workspace techniques by ID via the public API, with clearer authentication error messages for faster debugging

🛠 Fixes

• Multi-image @mentions now work correctly in Kling video generations

• Gemini image generations no longer fail on certain input assets

• FAUNA responses stream smoothly from the first word, and queued messages render in the right order

• Publishing techniques works reliably again

• Auto aspect ratio now correctly matches your input image's dimensions when running techniques

• Generated media frames size themselves to match the actual output

• Long ProRes exports complete instead of timing out

• PDF imports show proper page thumbnails in the library and publish flow

• Video nodes use the real dimensions of your footage for aspect ratio