DigiScribe

• 

  DigiScribe - Scholarly Annotated Transcripts

Inspiration

I wanted an easier way to preserve and quote text from inspiring Islamic audio content. I've struggled to find a software app that could handle both Arabic script and English language in transcripts (.srt). I've attempted to build a tool that would help make Islamic audio content accessible in text format authentically preserving its message.

What it does

✦ Transcribe Arabic and English Islamic audio and annotates areas that might need further review.
✦ Includes an audio playback with segment syncing to easily listen and compare to the text.
✦ Export to HTML or Google Docs for further editing and finalizing document.

How we built it

✦ Built with Google AI Studio.
✦ Gemini 3 Flash + Gemini 3 Pro with Extended Thinking.
✦ Prompt to code with support from Gemini Chat.

Challenges we ran into

✦ Setting up Google Docs export integration was difficult, but got it to work.

Accomplishments that we're proud of

It does more than transcribe. Each transcript segment includes the AI's complete "reasoning signature" - exposing internal logic checks, confidence assessments, and editorial notes. This transparency allows human reviewers to understand why the AI made specific decisions, not just what it transcribed. Reviewers can then approve, edit, or flag segments before publication.

What we learned

During testing with short duration audio files (approximately 2 mins) the output was excellent quality. After launching the app in the cloud and testing with 20 minute audio files it began to show truncated text. After a chat with Gemini about the issue it seems my app is making "One Giant AI Call". These limitations can be fixed by implementing a database and better architecture between Gemini 3 Flash and Gemini 3 Pro with Extended Thinking.

What's next for DigiScribe

Plan to move toward a production database architecture.

Built With

• react
• tailwind