Talk To It — proof transcript
==============================
Engine: faster-whisper 1.2.1, Whisper "base" model, int8, CPU (2-core VM)
Input:  a 26.0-second recording of a generated voice (Kokoro af_heart)
        reading a known script aloud
Time:   2.5 s model load + 3.9 s to transcribe (~6.7x real time)
Date:   2026-10-23

WHAT THE ENGINE HEARD (verbatim):
"Your AI runs out of credits in the middle of your project. What if it
could finish anyway? Start by answering a few straight questions. Your
AI turns the answers into a plan, checks its own work against them,
and fixes what fails. Another AI can pick up right where the last one
stopped without paying the build at all. Spend $5 once or subscribe
for the month if you actually like it."

WHAT THE SCRIPT SAID (the only difference that matters):
"...without paying the BILL at all"   [engine heard: "build"]
"Pay FIVE DOLLARS once..."           [engine wrote "$5" — same meaning]

Everything else: word-exact. Accuracy, stated plainly: one content word
wrong in ~78 — good enough to dictate a brief and skim-fix, not good
enough to transcribe testimony unattended.
