A religious content organization came to us with a genuinely hard problem: take the complete Sri Guru Granth Sahib — 1,500 pages of sacred Gurmukhi text — and turn every verse into a short, engaging social media video, paired with a second video explaining its meaning. The goal was educational outreach to younger generations who scroll Instagram and YouTube more than they read scripture.
Done manually, this was a months-or-years project. Done through commercial video-generation APIs, it would have cost thousands of dollars for the volume required. Neither option was viable, and the client still needed the sacred text handled with care — accurate translation, culturally appropriate visuals, nothing cut for speed.
The build: OCR that actually understands the text, not just reads it
The system starts by processing the source PDF page by page. Instead of generic OCR, we used Gemini AI specifically because of how it handles Gurmukhi script — it doesn't just extract characters, it semantically separates each original verse from its explanation, returning clean, structured data instead of a wall of unsorted text. A Google Sheets counter tracks progress across the run, so the pipeline can process the document in batches and pick back up exactly where it left off if it's interrupted.
Two video styles, one pipeline
Every verse produces two videos with deliberately different visual treatments: the original Gurmukhi verse gets a golden glow effect reflecting the sacred nature of the text, while the explanation video uses a clean, modern black-highlight style suited to educational content. Both are rendered locally with Python and FFmpeg — no per-video API cost, no cloud rendering bill that scales with volume.
What this actually delivered
- 100x faster processing — the full 1,500-page text went from a months-or-years manual project to hours
- Zero marginal cost per video — running locally eliminated the thousands of dollars commercial APIs would have charged at this volume
- 1–1.5 minutes to generate each video, versus what would have taken days by hand
- 100% semantic accuracy on the Gurmukhi verse/meaning separation
The organization's own words summed it up simply: what looked like an insurmountable challenge became as easy as uploading a PDF and pressing a button.
For the full technical breakdown, see the case study: Sacred Text Publishing →
Why this matters beyond religious content
The specific use case here is unusual, but the underlying pattern isn't: a large volume of structured source material, a repeatable transformation that would otherwise take enormous manual effort or expensive per-unit API costs, and a workflow built to run unattended at scale. Publishers, course creators, and content teams with large archives face the same math — this is what it looks like when that math gets solved properly.
If you're sitting on a large archive that needs the same kind of treatment, that's exactly what we map out on a free strategy call.
A2B AI Technologies builds custom AI automation, AI agents, and RAG-based systems for businesses that want measurable results, not demos. Explore our services →
Follow along: Instagram @a2bai.tech · a2b.services · info@a2b.services
