SMF·SUPERSCRIBE
Can a prompt replace Superscribe?
Voice dictation — business call notes + dictation
Exhibit tracking slip
Verdict
The desktop dictation half — a hotkey, streaming transcription, an LLM cleanup pass, pasted into whatever's focused — is a solved, one-sitting build with mature open-source equivalents to build from. The actual hero product is not that: live-transcribing phone calls on your existing number needs carrier call forwarding, a pooled telephony number, a real softphone integration, and a call-audio pipeline that can only be debugged against live calls — that's the half this build deliberately leaves out.
Exhibit A — The prompt
Received on30.07.2026Build a push-to-talk dictation tool for macOS: a global hotkey held down records the microphone, streamed in real time to a speech-to-text API, showing partial results in a small floating panel as you speak. On release, run one LLM cleanup pass over the final transcript — fixing punctuation, dropping filler words — then paste the result into whatever field has focus, restoring the previous clipboard contents afterward. If no API key is configured for the real-time transcription service, fall back to recording locally and transcribing with a local Whisper model once you stop talking — slower, but fully offline and free. Keep a history of recent transcripts in a simple list with copy buttons.
Do not build live word-by-word typing into the focused field (paste on release is the honest, achievable version), Windows support, or — most importantly — the phone-call capture product: forwarding calls from your existing number needs carrier-level call forwarding and a telephony integration that's real infrastructure, not a weekend feature, and pretending otherwise would be dishonest about the scope. If call notes matter, upload a recording by hand instead.Opening prefills the prompt — press enter to run it.
Exhibit B — What you lose
- B.1 call capture on your existing phone number
- B.2 auto-filing of dictations and calls to the right project or client
- B.3 a searchable workspace of transcripts with reports and CRM drafts
- B.4 Windows support and signed auto-updating installers
Prior art
Exhibit C — Why people still pay: telephony infrastructure/integrated workspace breadth
People pay for the phone rail — answering calls normally on the number they already have, no separate app or device — and for everything downstream landing pre-organized: transcripts matched to the right client and turned into searchable history and CRM-ready notes without manual filing.
Questions
Can I bring over my Superscribe dictation history?
There's no structured export of past dictations — this build starts its own local history from scratch, though nothing about your workflow otherwise changes.
Will it work on my phone?
The dictation tool itself is a macOS menu-bar utility tied to system-wide hotkeys, so no. And the phone-call capture feature this doesn't replace wouldn't run on the phone either — it works by forwarding calls to separate infrastructure.
What does it cost to run?
The real-time transcription API costs a small amount per use; falling back to the offline local model costs nothing beyond your own machine.
What's the one thing that doesn't survive the rebuild?
Calls on your existing phone number. That needs carrier call forwarding into a real telephony pipeline — genuine infrastructure this build deliberately doesn't attempt, not a missing feature that got cut for time.
Related tools
Receipt