Art Ai-Studio

A multimodal AI art creation platform that uses Gemini 3 to understand prompts and generate creative visual art in real time.

Comment

Inspiration

As a developer, I often lose time switching between building and visualizing ideas. When Google AI Studio launched, it showed how much faster creativity can move when you can iterate with AI in one place. That insight led me to build Art AI-Studio, a workflow tool for artists, filmmakers, and creators who rely on imagination but need a faster way to test and refine concepts. And also in order to spend less time on repetitive steps and more time creating.

What it does

Art AI-Studio is a two-track creative studio. Together, these tracks form an AI-powered pre-visualization workflow for filmmakers and artists, helping creators move from an idea or script to an editable storyboard and refined visuals in one place—so they can iterate faster and communicate concepts clearly. The product runs in two modes: Director Studio, which turns scripts into structured scene breakdowns and storyboard panels with per-panel edits (for movie or ad pre-production), and Artist Studio, which converts ideas into draft visuals and supports iterative artistic refinement. A key differentiator is annotation-based editing, Gemini 3 interprets your annotations and applies targeted visual updates for planning, storyboard drafting, creative iteration, and pre-visualization.

How we built it

*Tech Stack * 1.Frontend: React + TypeScript (Vite) 2.UI/Interaction: component-based workflow screens + an annotation layer 3.3D planning (optional): Three.js(mostly for UI design in some parts) via React Three Fiber for simple scene layout / composition for later purposes. 4.AI layer: Gemini 3 API via the Google GenAI SDK There is no backend and database and also user authentication in order to create MVP.

Architecture designed the AI layer as a clean AIClient interface with two implementations: LiveAIClient for calls Gemini 3 models. DemoAIClient to returns cached outputs (structured results) in order to avoid use of api during user design and interface testing.

Artificial intelligence Gemini 3 drives the core pipeline for tasks such as, Script reasoning [structured breakdown: converts raw script into storyboard-ready panel directives (description + caption)]. By using multimodal image generation + style control, generates panels from scene prompts with optional reference-image style guidance. By inspiration Annotation-based editing as one strong feature that users annotate a storyboard panel with an instruction, and Gemini 3 uses that contextual cue to regenerate only the specified element, rather than re-rendering the entire scene. And , as a contextual assistant, Gemini AI power chatbot helps refine prompts, keep consistency, and speed iteration.

Challenges we ran into

  • API quota/rate limits during tests but solved with Demo mode replay
  • Keeping consistent response art-style while keeping UI doesn’t break.
  • Managing multimodal assets (text + images) while keeping the UX fast and stable

Accomplishments that we're proud of

  • Built a complete pre-visualization pipeline: script/idea and then structured scenes after that storyboard panels to visual generation with iterative edits.
  • Delivered Gemini 3-powered reasoning for semantic parsing and structured breakdown (scenes, captions, panel prompts) to make outputs storyboard-ready.
  • Implemented annotation-based editing, where Gemini interprets user annotations and applies targeted updates for composition planning and creative iteration.

What we learned

  • Designing a clean client interface to swap AI backends safely
  • Building reliable demos: deterministic fixtures + graceful fallbacks matter
  • Iteration UX is the product: speed + clarity beats complexity

What's next for Art Ai-Studio

To improve our support for creator. In director mode , we wish to provide 3-D terrain and environmental condition and movie props that can help visualization. And for artist like comic and manga creator , we wish to upgrade our annotate edit function similar to "Clip studio Paint". In this way, they can do creation , testing and drawing and idea visualization in one place.

Share this project:

Updates