PulseDeck

Inspiration

Every idea deserves more than just slides — it deserves a story. We noticed that even with AI tools, people still spent hours making presentations look good, finding the right visuals, and recording voiceovers. So we built PulseDeck — where your decks don’t just show your story — they speak it.

What it does

PulseDeck transforms your ideas into fully designed, voice-narrated presentations — in seconds. You simply describe your topic, and PulseDeck: Writes clear, structured slide content using Gemini Pro Designs slides with smart layouts and visuals from Imagen 4 Narrates each slide using Gemini Flash TTS Saves everything to your Firestore for easy editing and playback The result? A ready-to-present deck — that talks.

How we built it

We built PulseDeck entirely on Google Cloud, using Google AI Studio for the frontend AI experience and Cloud Run for backend orchestration. Frontend (AI Layer – Google AI Studio): Built using Gemini models via the @google/genai SDK Used Gemini 2.5 Pro for content and layout generation Gemini Flash TTS to create natural, slide-by-slide voiceovers Imagen 4.0 to generate visuals and backgrounds for slides The UI runs completely on AI Studio integrations, connecting users directly to the Gemini ecosystem Backend (Cloud Infrastructure): FastAPI hosted on Google Cloud Run for RESTful deck management APIs Google Firestore for storing user decks and presentation metadata Google Cloud Storage (GCS) for audio, images, and template uploads IAM for signed URL generation and secure file handling The backend handles persistence, API routing, and scalability — while the AI logic stays on the client side

Challenges we ran into

Handling the 32MB upload limit on Cloud Run led us to integrate signed GCS uploads directly from frontend. Ensuring that slide narration, visuals, and text stayed in sync was a challenge in the early iterations.

Accomplishments that we're proud of

Built a fully working AI presentation system with real voice narration end-to-end. Integrated Gemini, Imagen, Firestore, and Cloud Run seamlessly into one cohesive product. Successfully deployed backend and frontend in sync — all within a span of 5 days.

What we learned

Building PulseDeck taught us just how powerful and developer-friendly the Google AI ecosystem is. Google AI Studio made it incredibly easy to experiment, iterate, and connect directly with Gemini models — we could generate slide text, visuals, and even voiceovers from a single unified workspace. The built-in tools for prompt tuning, multimodal inputs, and model testing drastically reduced our development cycle. On the backend, Google Cloud Run proved to be a game-changer. We deployed our FastAPI backend and scaled instantly without worrying about infrastructure. Combined with Firestore and Cloud Storage, it gave us a clean, serverless foundation for storing decks, files, and AI outputs — all with minimal setup. In short, we learned that with AI Studio and Cloud Run, powerful AI apps can be built fast, run efficiently, and scale effortlessly — turning bold ideas into working products within days, not weeks.

What's next for PulseDeck

AI Presenter Avatars: Let Gemini generate on-screen presenters who narrate your slides in real time. Adaptive Storytelling: Slides that automatically adjust tone, visuals, and flow based on audience or purpose. Collaboration Mode: Real-time team editing powered by Firestore listeners. Brand Intelligence: AI learns your company’s style, colors, and tone for perfectly on-brand decks. 1-Click Exports: Seamless integration with Google Slides, Docs, and YouTube for voice-over video exports. PulseDeck Pro: Analytics, custom voices, and enterprise branding for professionals and teams. Our vision is to make PulseDeck the “Notion for Presentations” — an AI workspace where ideas are designed, narrated, and shared automatically.

Built With

• cloud
• cloudrun
• firestore
• gemini
• run
• storage
• tts
• vertex
• voice