1 Year of Building a "Bring Your Own API Key" Transcription App — and Why It's Now an Automation Hub

Description and use

WhisperDirect, built by a solo iOS developer, moved from a subscription model to a one‑time purchase, leveraging the user’s OpenAI key and on‑device Apple Speech, NVIDIA Nemotron 3, and OCR. This eliminated server maintenance, reduced costs, and added automation via webhooks.

Problem

Running a backend for an AI transcription app as a solo developer is risky and costly; a subscription model forces continuous server maintenance even with minimal usage.

Outcome

The app now offers transcription, summarization, and automation without backend costs, relying on user‑supplied API keys and on‑device processing, making it viable for a solo developer.

Workflow

• 1. Design app to use user’s OpenAI API key directly for Whisper API calls.
• 2. Remove subscription model; make app a one‑time purchase.
• 3. Shift from GPU server to CPU and test multiple ASR engines.
• 4. Adopt Apple Speech on iOS 26 after its accuracy and limits improved.
• 5. Port on‑device modules (Apple Speech, NVIDIA Nemotron 3 diarization, OCR) into the app.
• 6. Add summarization/minutes using OpenAI, Gemini, or custom LLM endpoints.
• 7. Implement webhook integration to post results to external services like n8n, Slack, Google Sheets.
• 8. Provide 7‑day free trial and on‑device features free of charge.

Evidence

"The moment you go subscription, you can't stop."

Source

The app itself is a one-time purchase. No subscriptions.

Source

Apple Speech’s accuracy jumped, and the old limits and instability were gone.

Source

I built this for myself, honestly. You can POST transcription / summary / minutes results to an external endpoint...

Source