Make a Quiz From a Lecture Recording (Auto-Transcription Workflow)

TL;DR. Three-step workflow: (1) transcribe the lecture recording (Whisper, Otter, native YouTube captions), (2) paste the transcript into SimpleQuizMaker's [AI quiz generator](/ai-quiz-generator), (3) review and edit. 15-minute lecture → publishable quiz in ~10 minutes of teacher time.

Why this workflow matters

Recorded lectures (in-person, Zoom, YouTube, Loom) are everywhere. Most of them produce no follow-up retention work. Adding a 5-minute quiz turns the lecture from passive consumption into active recall.

The bottleneck used to be writing the quiz. AI tools have closed that gap.

Step 1 — Get a transcript

Pick the option that fits your source:

Option A — Already on YouTube

Use the auto-captions. Open the video → click the three-dot menu → “Open transcript.” Copy. (Or paste the URL directly into the YouTube quiz tool and skip this whole step.)

Option B — Local audio/video file

Use OpenAI's Whisper (free, runs locally with the whisper CLI, or via a hosted UI like MacWhisper). 95% accuracy on most lecture audio. ~1 minute of processing per minute of audio.

Option C — Live transcription during the lecture

Otter.ai (free tier: 300 min/month), Zoom's built-in transcription, or Microsoft Teams transcription. Lower accuracy than post-hoc Whisper but live.

Option D — Manual notes

If you took lecture notes, you can skip transcription. The notes themselves work as source material.

Step 2 — Generate the quiz

Paste the transcript into the quiz builder:

Generation takes 10–30 seconds.

Step 3 — Verify

Transcripts have errors. The AI inherits them. Common error classes:

A 2-minute pass against the actual recording catches 90% of these.

A 30-second time estimate

| Step | Time |

|---|---|

| Get transcript (YouTube auto-caption) | 30s |

| Paste, configure, generate | 1 min |

| Verify against source | 3–5 min |

| Edit / regenerate weak items | 2–3 min |

| Share link | 30s |

| Total | ~8–10 min |

For a 1-hour lecture this is the most efficient workflow we've found.

Use cases

Flipped classroom

Students watch the recorded lecture → take the auto-generated quiz before class. You see who engaged and which concepts confused them, then teach to that.

Make-up work

A student missed class. Hand them the recording + the quiz. They get accountability without the teacher manually writing make-up material.

Conference talk follow-up

Watched a great keynote? Generate a quiz on the talk to lock the key claims in. Far more durable than just enjoying it.

Student-generated quizzes

Have students record themselves explaining a concept (Feynman-style), then generate a quiz from their own transcript. They'll see exactly where their explanation had gaps. See Feynman technique explained, then generate a quiz from their own transcript. They'll see exactly where their explanation had gaps. See [Feynman technique explained](/blog/feynman-technique-explained).

Privacy considerations

If your lecture contains FERPA-protected student speech (e.g., a Zoom discussion where students answered out loud), don't use third-party transcription services that retain audio. Run Whisper locally instead.

Related reading

Why lecture-to-quiz is now feasible

Five years ago, turning a 60-minute lecture recording into a usable quiz required either manual transcription (1-2 hours), expensive transcription services, or the assumption that you remembered enough to author questions from memory. By 2026, transcription is a button click and AI quiz generation completes the pipeline.

The realistic workflow today:

Total time from lecture end to assigned quiz: ~10-15 minutes.

Transcription quality matters

The quiz is only as good as the transcript. Quality varies by:

Run a 30-second spot check on the transcript before sending to the quiz generator.

What questions transcripts produce well

What transcripts produce poorly

Optimizing the lecture for transcript-based quizzing

If you know in advance you'll generate quizzes from your recordings:

Privacy considerations

Recording lectures introduces consent questions:

For class lectures with student voices, often easier to record only your own audio (lapel mic, hidden cardioid) rather than the whole room.

Generate a quiz from your lecture transcript →