Neural Studio

• 

  Empty Canvas before adding any source.

• 

  Team members and Team Chat to allow multiple user collaboration.

• 

  The Creator studio functions.

• 

  AI Live Speech Mode: Where we can talk to it as a researcher or creator that responds accordindly, To ensure better user interaction.

• 

  The Researcher studio functions.

• 

  When you select a node(anything on canvas) and send a query.

• 

  Home Dashboard Page to select an existing or create a new Project Workspace

Neural Studio

Inspiration

The creative process is often fractured. Researchers get lost in tabs and PDFs, while creators struggle to turn messy notes into polished content. We wanted to build a workspace that feels like a "second brain" one that doesn't just store information but actively helps you process it.

Inspired by the multimodal capabilities of Gemini 3.0, we imagined a surface where text, images, video, and audio live together, and where the AI isn't just a chatbot, but a proactive collaborator that understands the spatial relationships between your ideas.

---

What it does

Neural Studio is a dual-mode infinite canvas: ** End To End Tools that support implementing any idea, bringing ideas to reality.**

🧠 Researcher Mode (Powered by Gemini 3.0 Pro)

• Deep Reasoning: Handles complex "Thinking" tasks like Ethics Reviews, Experimental Design, and Literature Synthesis.
• Multimodal Analysis: Drag and drop PDFs, videos, or images to get instant summaries, citations, and connection maps.
• Intelligent Clustering: Automatically organizes messy brainstorms into thematic clusters.

⚡ Creator Mode (Powered by Gemini 3.0 Flash)

• High-Speed Ideation: Generates viral hooks, script storyboards, and thumbnail concepts in milliseconds.
• Asset Generation: Creates podcast audio (TTS) and concept art directly on the canvas.

• Repurposing Engine: Instantly transforms a single research paper into:

  • Twitter/X thread
  • LinkedIn post
  • TikTok script

Both Modes

*** AI Live Option to express our ideas efficiently and get the best results. *** To implement best outcomes with dedicated studio methods.

---

How we built it

🖥 Frontend

• React + TypeScript + Vite for a performant, low-latency UI

🤖 Core AI Engine

We integrated the Google GenAI SDK to access the latest Gemini models:

• gemini-3-pro-preview → Deep reasoning, large thinking budgets (up to 4000 tokens), complex instruction following

• gemini-3-flash-preview → Low-latency chat and real-time suggestions

• gemini-2.5-flash-image → Asset and concept art generation

• gemini-2.5-flash-preview-tts → Podcast voice synthesis

🧩 The Canvas

• Custom node-based editor (no 3rd party canvas libraries)

• Built for spatial AI:

  • Custom spawning logic
  • "Flying" to collaborators
  • Semantic node connections

🔁 State Management

• Collaborative-ready state machine

• Handles:

  • App modes (Research ↔ Creator)
  • Context switching
  • Version control

---

Challenges we ran into

⚖ Model Selection

Choosing between Pro vs Flash was complex.

Solution: We created predefined Studio Tasks that auto-route to the optimal model:

Task Type	Model
Critique	Pro
Ethics Review	Pro
Literature Synthesis	Pro
Brainstorm	Flash
Ideation	Flash
Hook Generation	Flash

---

🧭 Spatial Context

Teaching AI to understand 2D canvas context required a custom system:

A context-assembly function that serializes:

• Node content
• Spatial relationships
• Selection groups
• Semantic proximity

into structured prompts for the model.

---

Accomplishments we're proud of

🎙 Podcast Feature

Drag a PDF onto the canvas → Two AI hosts (Kore and Fenrir) discuss it in real time.

🎨 Seamless Mode Switching

• Visual changes (color palettes, gradients)
• Functional changes (AI tools)
• Instant transitions

Feels like two powerful apps in one.

🧠 Thinking Models

Successfully implemented Gemini 3.0 Pro Thinking Budgets for deep critique and reasoning.

---

What's next for Neural Studio

Agents to work on assigned tasks (Both auto and Manual task assignment).

Gen AI Media Generation tools (For complete production tasks).

• WebSockets
• Yjs / Liveblocks integration

📦 Local-First Sync

• Offline-first workflows
• Conflict resolution

🤖 Agentic Workflows

AI that doesn't just generate text, but:

• Moves nodes
• Connects ideas
• Organizes clusters
• Cleans workspace
• Builds structures autonomously

---

Vision: Neural Studio is not just a tool — it's a cognitive environment. A place where thinking becomes visible, and ideas organize themselves.

Built With

• built-with-google-gemini-api-(gemini-3-pro
• gemini-3-flash)
• lucid-react
• lucide
• typescript
• vite