What's free and what costs $79?

Recording, file upload, and transcription are free forever. The AI summary that turns the transcript into a structured candidate writeup (skills, strengths, risks, next steps) is the paid feature, unlocked by a $79 one-time license. You can use the free tier as long as you want.

Is my data really private?

Yes. HireScribe runs entirely on your Mac or Windows PC. Audio files, transcripts, and summaries stay on your device. Once the app and models are installed, it works with the network off.

Is it legal to record interview calls?

It depends on the state. Thirteen US states require every party on the call to consent before you record; the rest follow a one-party rule. Because recruiting calls cross state lines, the safe practice is to disclose and get a clear yes on every call. HireScribe builds that consent step in and logs a timestamped attestation.

What platforms are supported?

macOS 11 or later (Apple Silicon or Intel; Apple Silicon recommended) and 64-bit Windows 10 or 11.

What are the system requirements?

Minimum: 8 GB RAM and about 5 GB of free disk. Recommended: 16 GB RAM and about 10 GB free on an SSD. On Apple Silicon, transcription is GPU-accelerated via MLX. Intel Macs and Windows run transcription and summarization on the CPU, so a stronger CPU helps and processing is slower.

Do I need to install anything else?

No. Everything is bundled with the app: Faster-Whisper / MLX Whisper for transcription, pyannote for speaker diarization, and llama.cpp with a Qwen3 GGUF model for summarization.

How long does processing take?

A typical 6–7 minute interview takes roughly 5–10 minutes end-to-end on Apple Silicon. Longer recordings scale roughly linearly. Intel Macs and Windows are slower because transcription runs on CPU.

Can I use OpenAI or Claude instead of local models?

Yes. You can bring your own OpenAI or Anthropic API key for summarization. Local stays the default, and transcription always runs on your device either way.

What audio and video formats are supported?

MP3, WAV, M4A, OGG, FLAC, WebM, and MP4. ffmpeg is bundled, so video files get their audio extracted automatically.

Can I export my data?

Yes — Markdown, PDF, ATS-ready JSON, and CSV. Copy into your ATS or archive however you like.

Do candidates have to install anything or join a bot?

No. You record on your side of the call. There's no meeting bot on the invite, no browser extension, and nothing for the candidate to install.

What happens to my recordings if I uninstall?

They stay on disk until you delete them. On macOS they live in ~/Library/Application Support/hire-scribe/, on Windows in %APPDATA%\hire-scribe\. Uninstalling the app does not touch those folders.

Can I customize the summary format?

Yes. HireScribe ships with 11 templates for different interview and meeting types (general, technical, behavioral, sales, executive, phone screen, hiring manager intake, stakeholder meeting, client discovery, debrief, and voice memo) and you can edit them or add your own.

Interview notes for recruiters

Structured interview notes for recruiters.
No bot in the call. Nothing leaves your laptop.

HireScribe records, transcribes, and summarizes candidate interviews into ATS-ready writeups on your own laptop. Six interview templates, speaker diarization, and exports built for Bullhorn, Greenhouse, and Lever. Made for boutique staffing firms, solo recruiters, and executive search.

Built for recruiters

Runs locally on your laptop

macOS full · Windows mic + file import

Free to download and try

The Problem

You can't be fully present when you're taking notes

The real cost of running interviews the old way.

Split attention

Typing notes pulls you out of the conversation. Candidates notice when you're half-listening, and your follow-ups get weaker.

Back-to-back blur

By the fifth interview of the day, details run together. The nuance that actually separates candidates is the first thing to fade.

Bots in the room

Meeting bots change how candidates show up. They also raise questions from legal and from candidates you'd rather not spend time answering.

Compliance exposure

Cloud transcription means interview audio leaves your laptop and lands on someone else's server. That's a real footprint to defend.

How it works

Three steps, all local

Record, transcribe, summarize on your own hardware, end to end.

Record

Start a recording before your call. A built-in consent gate prompts you to inform everyone on the line, and the timestamp is logged.

Works for phone screens, video calls, and in-person interviews.

Transcribe & diarize

Faster-Whisper or MLX Whisper converts audio to text on your device. pyannote separates speakers so you can tell who said what.

No cloud upload. Audio never leaves your laptop.

Summarize & export

llama.cpp runs a bundled Qwen3 model locally to draft a structured summary. Export to Markdown, PDF, ATS-ready JSON, or CSV.

Prefer hosted models? Bring your own OpenAI or Anthropic key. Local stays the default.

Stack: MLX Whisper · Faster-Whisper · pyannote · llama.cpp · Qwen3 GGUF. Everything bundled, everything local.

See it in action

From a messy call to a structured writeup

The transcript captures who said what. The AI summary turns it into an ATS-ready candidate writeup — all on your laptop.

Priya Sharma — Backend Engineer · Technical Interview

Transcript · speaker-labeled