Record or import a vocal lesson, transcribe it with your own MiMo key, and draft clear parent feedback with DeepSeek. Edit and confirm every lesson before anything is logged. Local-first and built for independent vocal teachers on Mac.
How did Astra change the scope or ambition of what you built?
Maker
Before the challenge, Vocal Lesson Log was an internal prototype for one teacher. Astra changed the goal to a public release another teacher could install, understand, and use confidently. It helped us review the whole workflow, simplify first-run setup, make every AI draft editable and teacher-confirmed, improve recovery when a task is interrupted, replace classroom material with synthetic demo data, and document the release clearly. We also published the source code, downloadable build, test results, and checksums so the project can be inspected and improved. The ambition is now a trustworthy, local-first tool that can grow with feedback from independent vocal teachers.
Report
Maker
📌
Hi Product Hunt — I’m a vocal teacher, and this project started with a very ordinary problem: after a full day of lessons, I still had to turn scattered classroom notes into clear messages for parents and keep the lesson count accurate.
Vocal Lesson Log is a Mac app for that workflow. A teacher can record or import a lesson, transcribe it with their own MiMo API key, create a concise feedback draft with their own DeepSeek key, edit every word, and confirm the lesson only when it is correct. The workspace stays on the teacher’s Mac, and the app can also be used manually before any AI keys are configured.
During the GPT-6 Astra Challenge, Astra helped us audit the public build and turn an internal prototype into something another teacher can install. It found failure cases around interrupted lesson commits, unsafe API-key handling, repeated transcription charges during retries, stale text overwrites, and private terms hidden inside an old privacy scanner. We rebuilt those paths, added first-run API setup and recovery, and created a clean public release with synthetic demo data.
The part I care about most is human review: an AI draft never becomes a lesson record by itself. The teacher remains responsible for what parents receive and when a lesson is counted.
I’d love feedback from teachers and makers on three things: Is the first-run setup clear? Does the draft preserve the useful teaching details? Which part of the post-lesson workflow still feels slow?