Record the lecture. Get the notes, the quiz and the cards.

What you get from one recording

A finished lecture produces four things from a single pass, and each of them is derived from the transcript rather than from a separate upload or a second AI action.

• The transcript itself, searchable, so you can find the ten seconds where the professor said what the exam covers
• Structured written notes, organised into headings rather than a wall of speech
• A multiple-choice practice quiz, with an explanation attached to every answer rather than just a score
• A flashcard deck, generated from the same material, that behaves like every other deck in Semora

Recording that survives your phone dying

Audio is captured in five-minute segments rather than as one long file, and the reason is specific. An .m4a killed before its writer finalises has no moov atom — the file is not shortened, it is unplayable. A single-file recorder that meets a low-battery shutdown or an aggressive iOS memory reclaim at minute 70 does not hand back 70 minutes. It hands back nothing.

Segmenting turns an unrecoverable loss into a survivable one: the worst case is the last few minutes, and every earlier segment has already uploaded. It also keeps each upload around 1.2 MB, which is what makes this work on campus wifi rather than only on a good connection.

The capture settings are tuned for one voice in a large room — mono, 32 kbps, speech-range sampling. A 90-minute lecture is roughly 22 MB. The same lecture at the audio presets a recorder ships with by default would be 86 MB.

The audio is deleted once the transcript exists

As soon as the transcript is durably written, the recording is deleted. This is not a storage-saving measure and it is not configurable — playback is deliberately not a feature.

The reasoning is that a lecture recording is not only your data. It contains your instructor's voice and everyone within microphone range of your phone, none of whom chose to be recorded by an app. Keeping that audio after it has served its purpose buys a hosting bill and an erasure obligation and nothing else. The transcript, the notes, the quiz and the cards are yours and they stay.

Permission comes first, and the app says so

Before your first recording Semora shows a sheet asking you to confirm you have permission, and pointing you at your instructor's rules and your school's policy. Many courses require permission before you record, some prohibit it, and the law varies by state and country.

That screen exists because an app that makes recording a one-tap action has some responsibility for the moment before the tap. It is not legal advice and it is not a substitute for asking — it is a prompt to ask.

Where it sits in the rest of Semora

Recording lives under Notes in the app, alongside uploads: slides, a chapter, or a photo of the board all go to the same place and produce the same set of outputs. A lecture is one input, not a separate product.

The cards generated from a lecture are ordinary Semora flashcards, so they open in the same reviewer as a deck built from a syllabus. The notes are attached to the course you recorded for, which means the AI Tutor can answer from them the same way it answers from a scanned syllabus.

The limits, plainly

One recording runs up to 90 minutes, and you are warned before the cap rather than cut off at it. Recording requires the phone app — iPhone or iPad — because it needs a microphone and a foreground audio session; the web app can read everything a recording produced, but it cannot capture one.

Semora checks for free space before it starts, because a device that fills up mid-lecture is the one failure with no recovery: the class does not happen twice. It also warns you if the battery is low and you are not plugged in.

On the free tier a lecture spends the account's single lifetime AI action, the same one a syllabus scan would spend. Recording more than one class is a Pro feature.

Frequently asked questions