Design options · Options from your own frames · Votes in team designs
Ask Velork AI for options and choose on the canvas: several takes on a design side by side, an options bar to step through them, compare and use one, with votes when you design with a team. Canvas panning is steadier, the canvas stays sharp after you move, and your Claude and ChatGPT plan limits are one hover away.
New
- Design options: ask for options ("give me 3 hero options") and Velork AI builds each as its own frame beside your design, then shows the options bar: step through them with the arrows (or ← →, ⌥[ ⌥]), Compare them side by side, and Use one (⌥↵): the others go and it takes the original's place, and ⌘Z brings them back. More like this, Another option and Make your own version are in the bar's menu; ⌘D on an option adds your own take
- Options from your own frames: select one frame for an Options chip that asks Velork AI for options on it, or several frames to make them options yourself
- Votes in team designs: in a team project, each person votes for the option they like and sees who else did; options, names and votes reach everyone in the live room
- Plan limits in the chat: hover the context ring in a Claude Code or Codex chat to see your 5-hour and weekly limits (monthly on free ChatGPT plans), how much is used and when each resets, with See all for the Spend page
Improvements
- Steadier panning: fast trackpad flicks and momentum no longer stall for a moment (some of their steps were taken for mouse-wheel notches), and panning is a little faster
- Sharp after moving: the canvas no longer stays blurry after a pan or zoom; what was shown soft while moving is drawn sharp again as soon as you stop
- Any device in the browser: the AI can test a page on the phone or tablet it names (iPhone 14, iPhone 16 Pro, Galaxy S23, iPad Air), at a breakpoint (sm, md, lg, xl) or at a size like 1024x768, instead of failing with "No device"; one it doesn't know opens as the closest kind of device. Desktop is the 1920 one again
- Live plan limits on the Spend page: Claude Code and Codex limits are read from the providers each time, instead of waiting for a chat to report them (they often showed nothing or a day-old number)
Fixes
- Canvas keeps drawing after an error: a passing drawing error no longer stops the canvas for the rest of the session (no Velork AI cursor, edits not shown until a restart); it picks up again by itself
- Chats keep their memory when tools change: switching mode (Develop, Plan, Design) or tools connecting after Velork starts no longer starts the AI over in a new session with only a summary of the chat; it goes on in its own session with the whole conversation
- Right model and cost under each answer: after a chat resumes its session or switches model, the line under an answer names the model that wrote it (it could show Sonnet for an Opus answer) and the cost of that answer alone, not the earlier turns' too