One Minute Memories

Preserve life’s moments in seconds: upload text, audio, photo or video. AI instantly tags, summarises and maps your memories into an interactive, searchable archive. Share your one minute memories.

Comment

Inspiration

"One Minute Memories" honours my late family friend Kerry’s 2013 vision of preserving life stories as digital keepsakes. He dreamed of turning journalistic interviews into family DVDs, and I was fascinated by the potential to search across an archive of real people's accounts of significant moments in time. Over a decade later, the technology to truly realize that vision is finally here. This project uses AI and the modern web to make memory-sharing instant, interactive, and enduring—a direct evolution of Kerry's original dream.

What it does

"One Minute Memories" is an AI-powered, crowd-sourced memory archive. Users can submit a memory as text, an image, a video, or an audio clip. Our AI pipeline then intelligently processes the submission to:

  • Moderate the content for safety.
  • Enrich it by generating a title, a summary, and contextual tags.
  • Analyze it to identify the location, date, category, and event type.
  • Connect it to other memories by discovering thematic and personal relationships.

[cite_start]The result is a rich, browsable "synaptic web" of human experience, where users can explore stories by keyword, theme, geography, or by following the dream-like connections from one memory to the next [cite: 20, 28, 41-43, 49, 56, 58, 62, 65, 73, 85].

How we built it

This project was a masterclass in rapid, AI-assisted development, but also required deep, iterative debugging to bring to life.

  • [cite_start]Front-End Scaffolding: We used Bolt.new to generate the initial React and Supabase application structure, which saved days of setup time[cite: 4, 12, 14, 28].
  • [cite_start]Back End and Storage: Supabase serves as the project's backbone, handling the PostgreSQL database, user authentication (in a future step), and, crucially, file storage for all multimedia memories[cite: 4, 12, 14, 28].
  • The Data Pipeline: To create a rich initial dataset, we built a custom data pipeline.
    1. [cite_start]Research: We used Perplexity AI to research and gather authentic, first-person memories about our seed topic (Glastonbury Festival) from across the web[cite: 20].
    2. [cite_start]Parsing: We built a dedicated Data Parser Tool (also using Bolt.new) to transform the raw JSON from Perplexity into a structured format for our application [cite: 28, 41-43, 49, 56, 58, 62].
    3. Seeding: We then created a front-end seeder in the main app to upload this curated data directly into the Supabase database.
  • AI Processing for User Submissions: This is the core of our intelligent system. [cite_start]We built a Supabase Edge Function that uses the Perplexity Sonar API to process all new user-submitted memories [cite: 41-43, 49, 56, 58, 62]. This serverless function performs a two-stage process: first, it moderates the content for safety, and second, it enriches the memory with a complete set of metadata, ensuring all user-generated content has the same quality and structure as our curated data.

Challenges we ran into

This was a true hackathon journey, filled with complex challenges that required persistence and creative problem-solving.

  • [cite_start]The Time Crunch: Building a project of this complexity as a solo founder under significant time constraints (and parent-on-duty duties!) was a major challenge[cite: 16].
  • The Unreliable Editor: Our biggest technical hurdle was the Supabase web editor, which was silently corrupting our Edge Function code every time we deployed it. We solved this by abandoning the editor and developing a more robust workflow: using Bolt to create the file locally and then using the Supabase CLI from the Bolt terminal to deploy it directly.
  • Debugging the Black Box: The Perplexity API is incredibly powerful, but getting the API calls right from within the Deno environment of a Supabase Function was a significant challenge. We had to methodically debug issues related to model names and the precise schema requirements for JSON mode, relying heavily on the function logs to guide us to the solution.

Accomplishments that we’re proud of

  • Building a Full End-to-End AI Pipeline: We successfully designed and implemented a sophisticated system for ingesting, moderating, enriching, and displaying user-generated content, making the platform truly scalable.
  • Creating the "Synaptic Web": The potential_relationships feature, which creates explorable links between memories, is a core part of the original vision that we successfully brought to life.
  • Solving Complex Technical Issues: Overcoming the challenges with the Supabase editor and the Perplexity API required deep debugging and a resilient, iterative approach. We are incredibly proud of the robust, working solution we engineered.
  • A Rich, Multimedia Experience: The platform now successfully handles text, images, video, and audio, making it a truly multimedia archive.

What we learned

  • AI-Assisted Development is a Superpower: Bolt.new proved to be an incredible accelerator for getting the project off the ground and for building new components quickly.
  • Serverless Debugging is a Skill: When working with Edge Functions, detailed logging is not optional—it's the most critical tool for understanding and solving issues.
  • Trust the Error, Not the Documentation: When a third-party API's behavior contradicts its documentation, the error log is your source of truth. The "Invalid model" error was the key that unlocked our final solution.

What’s next for One Minute Memories

With the core engine now built and proven, we are excited to:

  • Expand the Seeded Archive: Use our data pipeline to add memories from diverse global events, creating a richer and more representative map of human experience.
  • Enhance Data Visualization: Integrate Mapbox or Leaflet to create a geographical view of memories and explore more visual ways to represent the "synaptic web" of connections.
  • Build Community Features: Implement user registration, the ability for users to edit their own memories, and features for creating personal collections.
  • [cite_start]Develop B2B and Institutional Partnerships: Offer licensed access to anonymized datasets for cultural institutions, academic researchers, and media organizations, fulfilling the project's long-term ambition[cite: 23].

One Minute Memories began as a tribute to a friend's vision, and it has evolved into a powerful and practical experiment in emotional technology. [cite_start]By stitching fleeting moments into a rich multimedia patchwork, it offers a new way to connect with the past, understand the present, and create new narratives together[cite: 25].

Built With

  • beautiful-soup
  • bolt.new-managed-hosting
  • bolt.new-rapid-build-platform
  • elevenlabs-voice-api-(audio-summaries)
  • github
  • javascript
  • map-visualisation
  • openai-api-(tagging
  • public-web-endpoints-for-seeding
  • public-web-endpoints-for-seeding-**data-seeding-libraries**:-beautifulsoup
  • python
  • react-(via-bolt.new-scaffold)
  • scrapy
  • scrapy-**hosting-&-deployment**:-bolt.new-managed-hosting-**version-control-&-ci**:-github-**future-integrations**:-elevenlabs-voice-api-(audio-summaries)
  • summarisation)
  • supabase-(postgresql-+-storage-buckets)
Share this project:

Updates