This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
What I Built
LectureAI turns a lecture recording into everything you need to revise it. Upload an MP3, WAV or M4A in Hindi, English or Hinglish, and it gives you:
- a timestamped transcript
- a summary and a topic timeline
- likely exam topics, each linked to the moment it was taught
- structured notes
- flashcards with a revision queue
- assignments and deadlines the teacher mentions, collected on a calendar for you to confirm or dismiss
It saves the hour you would spend replaying a recording, pausing and typing notes. It also catches the "submit this by Friday" line you missed while writing something else. The lazy student gets to actually listen in class.
By default, everything runs on my own computer. The dashboard shows an "All local services ready" badge, and no audio or transcript is sent to the cloud. If you choose the optional ElevenLabs engine, that lecture's audio is uploaded for transcription.
Demo
🎥 Video demo: https://www.youtube.com/watch?v=dIBrWXM4P5E
Dashboard: courses, lectures, flashcards due and open deadlines at a glance.
Lecture view: summary, topic timeline and likely exam topics sit next to an audio player, so every topic is one click away from the audio.
Assignments: deadlines detected from lectures appear on a calendar with Open, Needs review, Done and Dismissed tabs. The AI proposes and you decide.
Code
https://github.com/AMAN240310/Lectur-to-everything
How I Built It
LectureAI is built on open-source models running locally, with an optional cloud engine:
- Whisper transcribes the audio, including Hindi and Hinglish, with timestamps.
- Gemma, served through Ollama, reads the transcript and produces the summary, topic timeline, exam topics, notes, flashcards and deadline candidates.
- ElevenLabs Scribe is an optional cloud transcription engine. Whisper runs locally by default, and you can switch a lecture to ElevenLabs when you want faster or more accurate transcripts and are fine with the audio being uploaded.
The pipeline is: upload → Whisper (or ElevenLabs) transcript → Gemma structured extraction → notes, flashcards, deadlines → dashboard.
A few design choices that mattered:
- Timestamps everywhere. Every topic and exam topic links back to the audio, so you can check the AI against what was actually said.
- Human in the loop for deadlines. Detected assignments go into a "Needs review" tab instead of silently landing on your calendar, because a wrong deadline is worse than a missing one.
- Reprocess and processing log. Any lecture can be re-run, and the log shows what happened at each step, so weak output is fixable instead of a dead end.
Why Does Open Innovation Matter?
A lecture recording is private: your teacher's words, your classmates' questions, your own voice. With a closed API, every recording is uploaded, billed per minute, and useless when the Wi-Fi drops. Open-weight models changed that:
- Privacy: by default, nothing leaves the laptop.
- Cost: a whole semester of lectures costs nothing beyond electricity, which matters for students.
- Language: Whisper handles Hindi and Hinglish, which many note-taking tools handle poorly.
- Control: I can swap model sizes, tune prompts and re-run processing as often as I like without worrying about a bill.
Prize Categories
Best Use of Gemma: Gemma, served locally through Ollama, is the brain of LectureAI. It turns a raw transcript into the summary, topic timeline, likely exam topics, notes, flashcards and deadline candidates, all on a laptop with no internet connection and no API bill.
Best Use of ElevenLabs: LectureAI can transcribe lectures with ElevenLabs Scribe as an alternative to local Whisper. The default path stays fully local, and ElevenLabs is an opt-in engine for people who want cloud-grade transcription.
The Lazy Payoff
In the demo, a 7-minute lecture became a full summary, a topic timeline, seven likely exam topics and flashcards with no typing on my side. Across four lectures, LectureAI generated 41 flashcards ready to revise.
Top comments (0)