iOS Stress Test — Transcription App Performance & Bug Report

I performed a sustained-use and usability review of Prof, an iOS transcription app using system-wide Broadcast. The test examined how the experience behaved over a longer session, including performance, controls, transcription output, and source changes.

Role: Manual QA Tester & Bug Analyst Device: iPhone 16 Pro Max Operating system: iOS 26.2 Network: Stable Wi-Fi 6 Duration: 60 minutes of continuous testing Report date: 15 January 2026

I reviewed long-session behaviour alongside the Broadcast status, Ask input flow, transcription delay, emoji rendering, multiple-speaker handling, audio-source switching, noise handling, and summary references.

The two-page report organized 11 issues by ID, category, severity, user impact, and recording timestamp: one critical, four high, five medium, and one low. Severity reflects the assessment in the original report.

Critical — Heat and responsiveness degraded during sustained use. The report recorded overheating after approximately 25 minutes and noticeable system-wide stuttering around the 50-minute mark, with evidence referenced at 00:52:15. These were session observations; the review did not establish an instrumented CPU/GPU root cause.

High — Battery use limited longer sessions. The report recorded an approximate 18–22% battery drop during the one-hour session. This is a result from that test environment, not a universal battery benchmark.

High — The Ask input flow became obstructed. At 00:22:45, the report referenced keyboard overlap and an unresponsive mic/speaker control. The recommendation was to resolve the input conflict and preserve access to the control while typing.

Other findings included missing Broadcast-status feedback, approximately 5–7 seconds of transcription delay, broken emoji rendering, merged speaker dialogue, missing separators after source changes, nonsensical text during ambient noise or silence, and summaries without transcript-linked timestamps.

The two pages below reformat the original issue log and six recommended improvements for readability. Issue counts, priorities, and timestamps are retained. Timestamps reference the original test recording; the recording itself is not embedded in this portfolio.

I delivered a developer-oriented report with issue IDs, priorities, impact, and timestamp references. Recommendations covered speaker diarization, timestamped citations, silence detection, source segmentation, performance profiling, and the Ask input conflict.

The demonstrated outcome is the completed test and actionable report. No post-fix validation or measured performance improvement is claimed.

Preparing an iOS release? Send your app, the workflows you want tested, session-length requirements, and deadline through Contra.