Back

Parrap

A realtime AI translation product for multilingual church services, built across web, mobile, desktop, backend, and live AI infrastructure.

Timeline
Nov. 2025 - current
My role
Solo builder and designer
Links
parrap.app
Realtime captions and operator dashboard

Overview

Parrap is a live AI translation product for multilingual church services. I designed and built the product end to end across web, mobile, desktop, and realtime AI infrastructure.

Why this problem mattered

Live church translation volunteer reference

Many multilingual churches in the U.S. rely on translation to help people worship together, but language barriers can still keep the community separated.

Most live translation depends on church volunteers, which makes recruiting difficult and quality dependent on preparation time, script availability, and each translator's language skill.

How God Made Me Happy in Him

Live

Speaker

English▾

Translation

Spanish▾

Live operator console

The operator surface brings captions, summaries, verses, audio status, and session controls into one place so the service can be monitored without exposing every technical detail.

  • Context buildingUpload sermon scripts, vocabulary, and church-specific terms before service so the AI has the context it needs for more accurate translation.
  • Easy QR code sharing with membersGenerate a QR code members can scan to join the session on their phone without account setup friction.
  • Multi-user access within one orgLet multiple operators or admins manage sessions within the same church organization.
  • Customizable display overlayControl what appears on the projection or sanctuary display without exposing the full operator console.

How God Made Me Happy in Him

CaptionsSummaryVerses

Buenos días, iglesia.

Es tan bueno verlos a todos aquí hoy,

y dar gracias a Dios por este día.

Hoy quiero hablar sobre lo que significa permanecer en Cristo.

Muchos de nosotros venimos agotados después de una semana larga.

Dios nos invita a descansar en su presencia antes de escuchar.

Member experience

Members can follow from web or mobile with live captions, translated audio, generated summaries, and detected Bible references without the church projecting translation for everyone.

  • Captions and audio availableFollow live with segmented captions, translated audio, or both depending on how they prefer to listen.
  • Periodic summarizationReceive rolling summaries of the sermon so late joiners can catch up without rereading every caption.
  • Bible verse detectionAutomatically surface detected scripture references so members can follow along in their preferred language.

Learnings

1.Trust came from reliability

People were willing to try AI translation only when the product felt stable enough to rely on during a real service. If captions failed, lagged, or disappeared, trust was difficult to regain.

2.Stable prototypes created better feedback

I had to be intentional about when I asked pastors and members to test the product. Releasing quickly mattered less than protecting confidence in the experience, because unreliable early tests could make people hesitant to keep using it.

3.Preferred terminology shaped belonging

Pastors cared about more than technically correct translations. They wanted familiar wording for church language, sermon terms, and repeated phrases so guests and newcomers could understand the service and feel part of the community.

4.Different stakeholders shaped different priorities

Pastors prioritized translation accuracy, members cared about how they experienced translation across text, audio, summaries, and Bible references, and operators needed clear controls for starting, pausing, clearing test captions, and managing audio inputs.

5.Implementation constraints became design material

The AI model's available inputs, output structure, latency, and context limits shaped the product as much as user needs did. Designing close to implementation helped turn those constraints into clearer flows, controls, and expectations.