Skip to content

Turning your study notes into MP3s and MP4s (plain version)

Not built yet — this page describes the plan. The full write-up, including exact commands, is in ../technical/study-audio.md.

WorkWingman's flashcards already read themselves aloud in your browser (see prep-studio.md) — that's a live voice that plays while the tab is open and disappears when you close it. This is different: it makes real files — MP3s and MP4s you can copy to your phone, plug into your car, or leave running on a TV. Both exist; this one just makes something you can carry around.

What this is

Your study material (the gap analysis, flashcards, and interview-prep notes WorkWingman already built for a job) gets turned into audio and video files you can take anywhere:

  • For your phone or car — an MP3 with a catchy background music track under a narrator reading your study notes. Meant to keep you alert while you drive or commute.
  • For falling asleep — two MP3 options: one with a gentle rain/thunderstorm sound bed under the narration, and one where the narration itself gently drifts between your left and right ear (like a slow, calming back-and-forth — think a white-noise machine, not anything clinical).
  • For your TV — an MP4 that's either a plain black screen or a looping rain video, playing one of the above audio tracks in the background.

Two important promises

This is not therapy and doesn't claim to fix anything. You might have heard of EMDR — a real, licensed therapy that uses side-to-side eye movement and is administered by trained professionals for trauma. The "gently drifts between your ears" sleep option is not that. It doesn't mention EMDR anywhere in its own copy for a reason: it's just alternating stereo sound, the same category as a fan or a white-noise app. It doesn't diagnose, treat, or cure anything. If you need help with anxiety, trauma, or sleep, please talk to a licensed professional — this is background audio for studying, nothing more.

The sleep versions are not for driving. They're built to help you relax and drift off, which is the opposite of what you want behind the wheel. The car version is the one built to keep you engaged and alert — that's the one for the road.

What you actually hear

The narration isn't just your notes read verbatim — it's written to work while you're only half paying attention:

  • "Explain it like I'm 5" for topics you're shaky on — plain-language analogies before any jargon (a hash map explained as a coat-check ticket, that kind of thing).
  • Pattern spotting for coding/interview prep — short, repeated lines like "when you see a problem about the longest stretch without repeats, that's a sliding-window problem," or "binary search is fast because it cuts the search in half every time." Repeating the same cue-to-pattern pairing a few different ways is intentional — it's built to sink in on a half-asleep or half-distracted listen, not to be understood in one pass.

Before any of that becomes audio, it gets fact-checked by a second AI. Claude writes the first draft of any technical claim (Big-O speed, which pattern applies, how a data structure trades off) — but a different AI (Codex, and optionally a second one if you've connected your own key) independently checks it against your own trusted study sources before it's allowed to become a narrated file. If the two disagree and it can't be resolved automatically, that piece is never presented as settled fact — it's either left out or clearly marked "unverified, please double check." The whole point: half-asleep or driving is the worst possible moment to accidentally memorize something wrong, so nothing technical reaches your ears without a second opinion first. This same check applies everywhere WorkWingman generates study material, not just audio — flashcards and Claude study projects get the same second look.

Where the sounds and voice come from

  • The narrator's voice is a free, offline text-to-speech program called Piper — it runs entirely on your own computer, reads your study text, and never sends anything over the internet.
  • The music, rain, and video clips are all either "CC0" (meaning: free for anyone to use, for anything, no permission needed) or generated fresh on your computer — never a copyrighted song or clip someone else owns. Every clip used gets its source and license written down so it's always clear where it came from.
  • Nothing is sent anywhere by default. All of the mixing and file-making happens on your machine. If you ever choose to plug in a paid, nicer-sounding voice or music service (entirely optional, and off unless you turn it on with your own account), only the text you're reading gets sent — never your resume or job details.

What you'll need installed

Two small, free programs — ffmpeg (which mixes and builds the audio/video files) and Piper (the narrator voice). WorkWingman already has a "dependency doctor" that checks what's installed on your machine and asks your permission before installing anything — these two just get added to that same checklist. Nothing installs itself without you clicking approve.

What's next

This page describes the design. The actual building work is split into a few focused pieces — the audio mixing, the video assembly, and the installer wiring — so they can be built and tested one at a time without stepping on each other. See the technical doc's build-agent breakdown for the full plan.