GoldenRetriever.ai Public Beta
Do Your Bit Ltd. · Video
GoldenRetriever.ai Public Beta is a macOS desktop app that turns your local media library into something you can actually query. It indexes videos, audio, PDFs, slides, screenshots and documents on your Mac, then lets you search by meaning instead of filename. The catch is that "meaning" here is literal: it reads what's on screen and hears what's said, not just the transcript. Your original files never leave your disk, and the AI work runs on your own Google Gemini API key.
About GoldenRetriever.ai Public Beta
What Is GoldenRetriever.ai Public Beta
Most search tools treat video and audio as opaque blobs. They either ignore them completely or run a speech-to-text pass and call it a day. GoldenRetriever.ai takes a different route: it uses multimodal AI to understand the original media directly. Slides on screen, a whiteboard diagram, the emphasis in someone's voice all get embedded and become searchable.
The product targets people drowning in recorded material. Filmmakers with years of footage, designers with saved screenshots, lawyers with hundreds of PDFs, anyone who's ever stared at a folder full of files and thought, "I know I saved that somewhere, and I have no idea what I named it."
The biggest limitation is the platform. This is macOS 14 or later only, and Windows and Linux are still on the roadmap. You also need your own Gemini API key, and those Google costs are billed separately from your GoldenRetriever subscription. The app is in public beta, so some polish is still missing.
Getting Started
- Download the beta build from the official site. It runs on macOS 14 or later, and the free tier needs no credit card.
- During onboarding, paste a Google Gemini API key. Embedding and AI traffic go straight from your Mac to Google on that key.
- Point the app at the folders you want indexed, including your video, audio, PDF, slide and image directories.
- Let the first index run. A bundled Qdrant vector database stores the embeddings locally on your machine.
- Search in plain English or ask a question. Answers come back with source files and timestamps.
Product Information
A quick look at GoldenRetriever.ai Public Beta's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Video editors and filmmakers
- Lawyers and researchers
- Designers and product teams
Tasks
- Finding a moment in long recordings
- Cross-format research
- Building a searchable archive
Scenarios
- Solo creators managing years of personal media on one Mac, where keeping files local matters more than cloud sync.
- Small studios and teams that want search across shared project assets but aren't ready for a hosted enterprise platform.
- Anyone with privacy constraints who prefers a local vector index and their own AI billing over handing media to a third-party service.
Key features
Multimodal Understanding of Video and Audio
This is the feature that separates GoldenRetriever.ai from transcript search. The AI processes visual content natively, so it catches a product demo or a whiteboard diagram that never got spoken aloud, which means a screen recording full of silent screen-share footage is still searchable by what appeared on the screen rather than by whatever was accidentally said over it. Audio is embedded as a full signal, which preserves tone and emphasis a text pass would flatten.
Semantic Search With Hybrid Ranking
You describe what you remember, not what the file is called. The hybrid search engine blends semantic vectors with BM25 keyword ranking, so a loose description and an exact phrase both surface the right file even when your memory of the content is fuzzy and incomplete. That mix matters when you only recall part of a name or a rough idea of the content.
AI Q&A With Sources and Timestamps
Ask a question and the app answers using your own library as evidence. Every answer points back to the source files with timestamps, so you can verify the claim instead of trusting it. This is the difference between a chatbot guess and a citation you can open.
Local Vector Database
Embeddings and the search index live on your Mac inside a bundled Qdrant binary. Nothing about your content gets copied to GoldenRetriever's servers for storage. If you handle sensitive material, keeping the index local is a real advantage over cloud-only tools.
Bring Your Own AI Key
You paste a Google Gemini API key during setup, and AI traffic goes directly from your Mac to Google. GoldenRetriever doesn't see the key and doesn't mark up the bill. It also means your compute cost scales with how much you actually index, not with a flat credit bundle.
Flexible Q&A Model Choice
Gemini embeddings handle indexing, but for answering questions you can swap in Gemini, OpenAI, Anthropic, Ollama or any OpenAI-compatible endpoint. Teams already standardised on one provider can keep using it, and local Ollama setups stay fully offline.
Automatic Folder Watching
Once you point the app at a directory, it watches for new files and indexes them as they arrive. Recordings added next week get picked up without a manual re-scan. For people who record constantly, that removes the chore of remembering to re-index.
Pros and cons
Pros
- Understands video and audio content directly, so on-screen context and tone survive instead of getting lost in a transcript.
- Files stay on your Mac and the vector index runs locally, which suits privacy-sensitive work.
- Answers cite sources with timestamps, so you can check the evidence quickly.
- Free tier covers the full AI pipeline for up to 100 files, with no credit card needed.
- Beta pricing is locked for life, and you can choose your own Q&A model.
Cons
- macOS 14+ only. Windows and Linux users are stuck waiting, and there's no mobile app.
- You need your own Gemini API key, and Google bills you separately, so the true cost is higher than the subscription alone.
- It's public beta, so admin features, SSO and at-rest encryption are still roadmap items rather than shipping features.
Frequently asked questions
You describe the thing you remember, and the app returns matching files instead of exact-name hits. A media library search across years of footage usually finds the clip faster than scrolling through folders would.
Related content
Explore related tools, skills, and articles for GoldenRetriever.ai Public Beta.
GoldenRetriever.ai Public Beta Alternatives
ngram 2.0
ngram · Video · Marketingngram 2.0 is an AI video generator for teams. It turns a prompt, a screen recording, or a product URL into a narrated explainer, product demo, or AI avatar video, and every scene stays editable.
OpenScreen 2
OpenScreen (open-source project) · Video · ProductivityOpenScreen 2 is a free, open-source screen recorder and demo video editor for Windows, macOS, and Linux, with automatic zooms, styled frames, local captions, and agent edits.
Skreno
Skreno · Video · ProductivitySkreno is a browser screen recorder that captures a tab, a window or your whole display, then attaches AI transcripts, summaries and console logs to one share link.