From a feeling to an artboard
VIBE takes a description of a mood, an audience, and a purpose, then produces real HTML artboards rather than images. Because the output is markup, text stays selectable, layout reflows, and you can open the source and change a colour or a headline without regenerating anything. The supported formats are pitch decks, posters, landing pages, stories, cards, and invitations.
Why HTML instead of image generation
An image model gives you a picture of a layout; it cannot give you a layout. Text inside a generated image is frequently malformed, cannot be edited, cannot be translated, and cannot be made accessible. Emitting HTML means the result is a document: readable by a screen reader, editable in place, and copy-pasteable into a real page.
Direct source editing
Every artboard exposes its own source. You can adjust spacing, swap a palette, rewrite copy, or delete a section, then present the result full screen. This matters because the first generation is rarely final, and regenerating from scratch tends to lose the parts you liked.
Which models it uses
VIBE reuses whatever models you have equipped in the Model Organizer, so there is no separate configuration and no separate billing relationship. Stronger instruction-following models generally produce better structured layouts; smaller local models work but tend toward simpler compositions.
Frequently asked questions
Does VIBE generate images?
No. It generates HTML artboards, so the text remains real, selectable, editable text rather than pixels.
Can I edit what it produces?
Yes. Each artboard exposes its HTML source for direct editing, and you can present the result full screen when you are finished.
Which models does it use?
The same models you have already equipped in the workspace Model Organizer, billed against your own keys.
What formats are supported?
Pitch decks, posters, landing pages, stories, cards, and invitations.