Memoria

Product Introduction

1. Definition: Memoria is a local, on-device AI-powered search engine for personal photo and video libraries. Technically, it is a mobile application (iOS/Android) that performs offline indexing and semantic search across visual and auditory media content.
2. Core Value Proposition: It exists to solve the problem of digital memory hoarding by making every photo, screenshot, and video instantly searchable by their content—text, spoken words, faces, and objects—without compromising user privacy through cloud uploads or subscriptions.

Main Features

1. On-Device Optical Character Recognition (OCR): Memoria scans and extracts text directly from images and video frames stored on your phone. It supports 7 languages (English, French, German, Spanish, Italian, Portuguese, Japanese), enabling searches within documents, whiteboards, screenshots, and memes. How it works: The app uses local machine learning models to detect and transcribe text without an internet connection.
2. Local Audio Transcription with Whisper: The app transcribes speech from videos and audio files using OpenAI's Whisper model running entirely on the device. It supports over 99 languages, indexing lectures, meetings, and voice notes for search. How it works: The Whisper model is downloaded once (optionally) and then processes audio locally, converting speech to searchable text without data leaving the phone.
3. Private Face Detection and Clustering: Memoria automatically detects human faces in your media and groups them into distinct clusters (e.g., "Emma"). How it works: Using on-device computer vision, it identifies facial features and creates private, local groupings, allowing you to find all photos of a specific person with a single search query.
4. On-Device Object Recognition: The app identifies and indexes objects within photos (e.g., "red bicycle," "helmet"). This feature uses local AI models to tag visual content, enabling search by scene description without manual tagging.

Problems Solved

1. Pain Point: The "digital shoebox" problem—users have thousands of photos and videos but cannot find specific moments because they are buried in a chronological feed and lack metadata. Traditional search fails for content within images (text, objects) and videos (speech).
2. Target Audience: This includes students and professionals needing to find notes from whiteboard photos or lecture recordings; individuals with large personal media libraries; privacy-conscious users wary of cloud photo services; and anyone who frequently saves reference images, memes, or documents to their camera roll.
3. Use Cases: Finding a specific slide from a recorded lecture by searching for a spoken term; locating a photo of a receipt or document by searching for text on it; gathering all photos of a family member for a gift without manual scrolling; retrieving a saved meme or screenshot using a phrase written in it.

Unique Advantages

1. Differentiation: Unlike cloud-based photo apps (Google Photos, Apple Photos) which may offer similar search features but require uploading data to servers, Memoria performs all processing locally. It also differs from subscription-based services by being a one-time purchase (for the Plus tier) with no recurring fees, as it has no server infrastructure costs.
2. Key Innovation: Its fully local, privacy-by-design architecture is its core innovation. By running state-of-the-art AI models like Whisper for transcription and advanced OCR/vision models directly on the smartphone, it delivers powerful semantic search while guaranteeing that sensitive personal media never leaves the user's device.

Frequently Asked Questions (FAQ)

1. Is Memoria secure and private? Where is my data processed? Yes, Memoria is designed for maximum privacy. All data processing—including OCR, speech transcription, and face recognition—occurs locally on your iPhone or Android device. No photos, videos, or audio files are ever uploaded to the cloud or external servers.
2. How does Memoria's pricing work? Is there a subscription? Memoria offers a free tier with core features. Memoria Plus is a one-time purchase (not a subscription) that unlocks unlimited media indexing and priority background processing. The developer emphasizes a one-time fee is possible because the app has no ongoing server costs.
3. What languages are supported for text and audio search? For text-in-images (OCR), Memoria supports 7 languages: English, French, German, Spanish, Italian, Portuguese, and Japanese. For audio transcription in videos, it supports 99+ languages via the on-device Whisper model.
4. How long does initial indexing take, and does it drain battery? Indexing time depends on your library size (e.g., a few minutes for 1,000 items). It is designed to run efficiently in the background, even when the app is closed. The Plus tier offers "priority indexing while charging" to optimize for battery life.
5. Can Memoria search inside PDFs or other document files? Based on the provided description, Memoria can index text from PDFs that have been scanned or saved as images to your camera roll, as this utilizes its OCR capability. It primarily focuses on media within your phone's photo library.