Wonderful pdf

An AI studio that transforms Scripture into visual PDF storybooks, stored durably on Backblaze B2 with Genblaze SHA-256 provenance manifests and a cloud remix studio.

Comment

Inspiration

Bible Scripture contains timeless wisdom, powerful narratives, and foundational truths. To complement text-based reading, visual representation provides an extraordinarily impactful medium for learning, memory, and deeper reflection.

Scientific research highlights the immense impact of pictorial learning:

  • 65% of the population are visual learners.
  • The human brain processes visual information up to 60,000 times faster than plain text.
  • Studies show that people retain approximately 80% of what they see, compared to just 20% of what they read alone.

Driven by these insights, we built Wonderful PDF (Gospel AI Comic & Storybook Studio)—a platform designed to convert Scripture narratives into clear, vibrant, multi-panel visual storybooks. By bringing Scripture into a visual format, Wonderful PDF helps people engage with biblical stories through the proven power of pictorial learning.

What it does

Wonderful PDF is an AI-powered media creation platform that converts Scripture concepts into visual storybooks with full cryptographic provenance:

  1. AI Storyboard Planning: Takes Scripture text or Gospel concepts and uses LLMs (gpt-5.4-nano) to plan sequential comic scene panels with visual prompts and narrative captions.
  2. AI Panel Illustration: Uses gpt-image-2 to render high-detail, vibrant Christian illustration panels for every scene.
  3. PDF Storybook Assembly: Stitches the AI panels, titles, and narrative captions into a formatted multi-page PDF document ("Wonderful PDF") using ReportLab and Pillow.
  4. Durable B2 Storage & Genblaze Provenance: Saves every generated PNG panel, compiled PDF, and SHA-256 provenance manifest directly into Backblaze B2 Cloud Storage.
  5. B2 Vault & Remix Studio: Allows users to log in, view their saved works, select past B2 cloud images, and remix/re-order them into brand-new compiled PDF storybooks.

How we built it

  • Backend: Python & Flask with Tidb SQL (app.db) for user accounts, image history, and storybook management.
  • AI Processing: Azure OpenAI Service (gpt-5.4-nano for storyboard planning and gpt-image-2 for image panel rendering).
  • Parallel Concurrency: Used Python's concurrent.futures.ThreadPoolExecutor to render all AI image panels simultaneously, cutting processing times down to ~15 seconds.
  • PDF Engine: ReportLab and Pillow (PIL) for fast, pure-Python multi-page PDF compilation.
  • Cloud Storage & Provenance: Integrated Backblaze B2 (b2sdk.v2) and the Genblaze SDK (genblaze-core, genblaze-s3) to generate canonical SHA-256 provenance JSON manifests and host assets on s3.ca-east-006.backblazeb2.com.

Challenges we ran into

  • API Timeout Bottlenecks: Rendering 4 high-res AI images sequentially was causing 90+ second request timeouts. We solved this by implementing multi-threaded parallel execution (ThreadPoolExecutor), fetching all panel images concurrently in under 20 seconds.
  • Serverless Deployment Compatibility: Ensuring 100% compatibility with serverless platforms like Vercel by utilizing pure-Python libraries (ReportLab + Pillow) and eliminating system binaries like ffmpeg.
  • B2 Storage Authorization: Resolving private bucket access by generating authorized download URLs with Genblaze tokens so users can read PDFs and view JSON manifests seamlessly in their browsers.

Accomplishments that we're proud of

  • Sub-20-Second Pipeline: Building an end-to-end pipeline that turns a simple text concept into a multi-page PDF storybook in under 20 seconds.
  • Tamper-Evident Media Provenance: Successfully implementing Genblaze SHA-256 manifests to cryptographically verify every storybook generated.
  • The B2 Remix Studio: Giving users the ability to select past AI image panels stored in their Backblaze B2 Vault and re-stitch them into new storybooks.

What we learned

  • How to leverage Backblaze B2 and Genblaze for structured object storage and verifiable media provenance.
  • Techniques for parallelizing AI image generation requests to optimize performance for web applications.
  • Building clean, serverless-friendly Python pipelines using ReportLab and Pillow.

What's next for Wonderful PDF: Launching and Collaboration

  • Multilingual Scripture Support: Expanding the LLM storyboard planner to auto-translate Gospel storybooks into Spanish, French, Swahili, and other global languages.
  • Audio Narration Integration: Adding TTS audio tracks (via Genblaze Audio adapters) so pages can be listened to as read-along audiobooks.
  • Church & Creator Collaboration: Launching public sharing tools so Sunday school teachers, ministries, and creators can collaborate, curate, and publish open-access Gospel comic libraries hosted on Backblaze B2.

Built With

Share this project:

Updates

Submission history