Android app · version 0.1.0

Your lectures and meetings, transcribed and summarized offline.

LectureMind records or imports audio, transcribes it with Whisper and writes notes, minutes and summaries with a language model that runs on your phone. No servers, no accounts and no permission to go online: your audio never leaves the device.

Download APK · 9.9 MB All versions

Android 10 or later · 64-bit ARM · free

Lecture notes

Topic

Mechanical waves and how they travel through different media.

Notes

Speed depends on the medium, not on the source [3:12]
Wavelength = speed / frequency [12:30]
Example: sound travels ~4 times faster in water [27:05]

Review

What changes when a wave enters a new medium?

27:05

Illustration of a lecture's notes. The app's interface is in Spanish.

What it does

Record or import

Records with the screen off, pauses from the notification and recovers the recording if the app is closed. Imports MP3, WAV, M4A, AAC, FLAC, OGG and Opus.

Transcribe with Whisper

Spanish and English, with a timestamp for every sentence. If interrupted, it picks up where it left off. An hour of audio takes about 25–30 minutes on a mid-range phone.

Summarize by type

Lecture: topic, notes, concepts and review questions. Meeting: tasks, decisions and open items, without inventing owners or dates. Long recordings are summarized in parts.

Search and listen

Search a word across all your transcripts. Tap a segment or a timestamp in the summary and the audio jumps right there.

Export

To TXT, Markdown, PDF, SRT subtitles or JSON, in the folder you choose.

All on the phone

The AI models run on the phone's processor. It works in airplane mode.

Install

Download the APK

Use the Download APK button on your phone. Android will ask for permission to install apps from your browser or file manager: allow it just for this installation.

Get the AI models

They are not bundled (they weigh hundreds of MB) and the app never goes online to fetch them. Download them, then import them under Configuración (Settings):

For	File	Size
Transcription (recommended)	ggml-small-q5_1.bin	190 MB
Transcription (fast)	ggml-base-q5_1.bin	60 MB
Summaries (recommended)	gemma-3-4b-it-Q4_0.gguf	2.4 GB
Summaries (light)	Qwen3-1.7B-Q4_0.gguf	1.1 GB

The app checks each model's SHA-256 fingerprint when importing it. Gemma is distributed under the Gemma terms of use; Qwen3 and Whisper under Apache 2.0 and MIT.

Record, transcribe and summarize

Pick a category (Clase for lectures, Reunión for meetings, or General), tap Transcribir and then Resumir. Both tasks keep running with the screen off.

Requirements

Android 10 or later, a 64-bit ARM processor with the dotprod, i8mm and fp16 extensions (generally 2022 chips or newer). Without them the app records, imports and exports, but can't transcribe or summarize.
8 GB of RAM or more to summarize with Gemma 3 4B; with less, use Qwen3 1.7B.
Space for the models (250 MB to 3 GB) and about 115 MB free per hour of audio while transcribing.
The app's interface is in Spanish; transcription and summaries work in Spanish and English.

Tested on a Redmi Note 14 Pro+ 5G (Snapdragon 7s Gen 3, 12 GB of RAM). This is an early version: summaries are written by a small model and may contain mistakes, so check them against the transcript.

Privacy

LectureMind has no Internet permission: Android won't let it connect. There is no telemetry, no account and no server. Recordings, transcripts and summaries stay in the app's private storage and only leave the phone if you export them.

Recording other people may require their consent under your local laws: let participants know before you record.

Verify the download

SHA-256 fingerprint of the version 0.1.0 APK:

ab0eb247de6dadafcc4f25879d9f7213364d6f25555e7ace92eddd447295c395

On a computer: sha256sum LectureMind.apk.