Voice as the primary input
The interface treats speaking as the default interaction rather than an attachment to a text form.
01
Yappper
Voice-first research, shipped at hackathon speed
Next.jsGeminiPolarClerkSupabase
Overview
Yappper lets researchers, product teams, recruiters, educators, and creators collect richer feedback through text or voice. Respondents can open a shared link without creating an account, speak naturally, and leave survey owners with playable audio, AI-generated transcripts, and exportable responses.
Raenest × DevCareer · April 2026 · $3,000 prize pool
Ask with structure.
Listen with context.
01 / Challenge
The conditions that shaped the system.
A useful architecture begins with the real constraint.
Build an end-to-end AI product in three days without reducing it to a transcription demo. The result needed frictionless survey participation, useful review and export tools, secure accounts, persistent media, multilingual transcription, and a credible path to paid usage.
02 / System map
From input to outcome
01Shared survey link
→02Text or voice response
→03Supabase storage
→04Gemini transcription
→05Response review
→06Audio & data export
03 / Approach
How the build was organised.
Make the boundaries clear, then move.
- 01
Designed hybrid questions and a respondent journey that makes voice feel as natural as typing, with no account required to answer a shared survey.
- 02
Used Gemini for multilingual speech-to-text transcription, Supabase for audio and response storage, Clerk for authentication, and Polar for payments.
- 03
Focused the three-day build on one complete loop: create, share, collect, transcribe, review, and export.
Decisions that carried the product.
Managed infrastructure
Clerk, Polar, and Supabase reduced setup time while preserving production-shaped authentication, billing, and storage.
Small complete loop
The build focused on creating a survey, capturing a response, transcribing it, and reviewing the result.
What changed after the work.
- ↗Built and shipped a complete voice-survey product in three days.
- ↗Selected as one of ten Raenest × DevCareer hackathon winners in April 2026.
- ↗Turned audio responses into searchable transcripts while retaining the original voice and exportable data.
