Find · Meetings · Memory · Create — one private app · 7-day Pro trial

Find any moment
in everything you've recorded.

MediaFind transcribes, understands and indexes your video, audio and photos on your own computer — then lets you search them the way you remember them. Plain language in, the exact timestamp out. The same private library runs your meetings, keeps your AI chats and notes, and turns what you find into a video.

7-day Pro trial · monthly or annual · cancel anytime · see search modes · try the demo
macOS Windows (beta) Linux (beta) Signed & notarized on macOS
MediaFind — Library

Open models — running entirely on your machine

Whisper speech recognition CLIP visual search Local language model On-device embeddings Speaker diarization Voice recognition OCR Face recognition
100%
Runs on-device
0
Accounts & API keys
15+
Ways to search & recall
$8.99
Pro per month, billed annually
See it in action

Find in action, in under a minute

A real screen recording — search, Ask with cited answers, on-screen text, faces & celebrities, Knowledge & Insights, and jump-to-the-moment playback.

How it works

From a folder of files to a searchable mind

No pipelines to wire up, no cloud buckets, no keys. Point MediaFind at a folder and it does the rest — locally.

1

Add your media

Add a folder of video, audio or photos, or bring in chats and notes. MediaFind handles the common formats and, while the app is open, keeps watching your folders: new files are indexed once they've finished copying. You can turn that off in Settings.

2

It understands it locally

Transcripts, speakers, embeddings, keyframes, on-screen text, objects and chapters are computed on your computer — nothing is uploaded.

3

Find it, then use it

Describe a moment, ask a question, jump to the exact timestamp — then turn it into meeting notes, a reel or a new video, or hand it to your editor.

Find

Search it fifteen different ways

You don't always remember the words. MediaFind indexes meaning, sound and pixels — across video, audio and photos — so you can search by what was said, what was shown, who was there, or what it felt like.

💬 Natural language by meaning 🖼️ Visual scene CLIP 🔤 On-screen text OCR 🎵 Music & songs 🎨 Color & mood 😊 Emotion 🔊 Sound events 🗣️ Phonetic sounds-like 🏃 Actions 🧩 Named entities 🖻 Reverse image 🔗 Related clips 👤 Faces & people Pro 🎙️ Speakers 🅻 Brand logos Pro

Every mode is keyless and on-device. Results link to the exact timestamp — confirm in one click.

💬

Natural-language search

Describe a moment in plain words and get ranked clips with exact timestamps — semantic, not keyword. Jump straight to it, or cut it out as a clip of its own.

🔍 “the part where she explains the pricing change” → 3 results, top match 02:41
🧠

On-device AI assistant

Ask questions of your whole library and get grounded, cited answers — powered by a language model that runs locally. Multi-turn, conversational, and it never phones home.

“What did we decide about the launch date?” → answer with sources & timestamps
🖼️

Photos and frames, not just speech

Photos — JPG, PNG, HEIC and more — are indexed next to your video and audio. Search the picture by scene and objects, by the text in it, by color, or with a sample image.

🔍 “whiteboard with the Q3 roadmap” → a photo and two video frames
NEW
🗣️

Speakers & voices

Keyless diarization labels who said what, and each voice is matched on-device against voices you've named and a bundled gallery of well-known ones — so a name given once follows that voice into every podcast and voice-over. Unsure matches stay unnamed.

🎙️ Speaker 2 → ⭐ matches a known voice · ✎ confirm · ✕ dismiss
📝

Summaries & chapters

Per-file summaries and auto-segmented chapters so long recordings are skimmable in seconds.

Transcription on your machine

Whisper with word-level timestamps, and a names & terms list that keeps people and products spelled right. A Turbo tier brings near-Large accuracy to 8 GB Macs, and on Apple Silicon you can switch to the whisper.cpp engine, which runs on the GPU.

🧩

Every mention of a name

People, places and organizations are picked out of every transcript, so you can find each time someone or something came up.

PRO
👥

People & Faces

On-device face library — click a face to jump to every appearance, or switch it off and wipe all face data anytime. Famous public figures auto-named ⭐.

PRO
🅻

Brand-logo detection

Find where a brand appears — by name (“Nike”) or by an uploaded sample logo image.

PRO
🎞️

Highlights & reels

Stitch a file's best moments into one reel, or make a reel of every moment about a topic or tag across your library.

Meetings

Your meetings, written up on your computer

Record a meeting as it happens or process any recording afterwards. MediaFind writes the transcript with speaker labels, then pulls out action items, decisions and key points — with no bot in the call and nothing uploaded.

PRO
📋

Meetings mode

Transcribe live from your microphone, a browser tab's audio or — in the Mac app — both sides of a Zoom, Teams or Meet call. Key points, action items and decisions build as people talk, voices you've named are recognised next time, and any recording you already have becomes a meeting too.

🎙️ Live · Speaker 2 → Priya · ✅ “I'll send the budget by Friday”
NEW
📄

Word, PDF & calendar

Save any transcript as a Word document, PDF, SRT, VTT or text, written on your machine with nothing uploaded. With Pro, meeting notes export too — overview, action items and decisions up front — and action items go to your calendar as an .ics file.

Memory

Your AI chats and notes, searchable and yours

What you've worked through with ChatGPT or Claude, and what's in your notes, can sit next to your recordings — on your computer, not in another company's cloud memory. Search it, ask it, and forget any of it.

💭

Chats and notes, kept here

Import a ChatGPT or Claude data export, a text file or a pasted chat, or a folder of Markdown notes such as an Obsidian vault. Importing again skips what hasn't changed, and chats and notes don't count toward the free plan's 10 files.

📥 conversations.json · 📁 an Obsidian vault → your “personal” library
🔎

Ask it, then forget it

Ask a question and the answer lists the messages it read — or says your conversations don't cover it rather than guessing. Chats turn up in Find results too. Export a library, delete a conversation or reset a whole library whenever you like.

“When did I say the lease renews?” → answer + the messages it looked at

AI assistants you connect over MCP can save to and search the same memory. Set it up →

Create

Turn what you find into a video

Create is a timeline editor built into MediaFind. Search your library from its side panel, drop clips on the timeline and cut — without exporting anything first.

🎞️

Timeline editor

A multi-track timeline with split, dissolves and undo. Cut by editing the transcript — drop silences and filler words, or keep one speaker — burn in captions, add titles, reframe for vertical, square or wide, and render MP4 or ProRes.

✂️

Built-in video tools

Keyless ffmpeg editor: cut, convert, resize, compress, extract the audio, grab a frame, or make a GIF — without leaving the app.

PRO
🗂️

Compose storyboard

Describe the video you want, or paste a narration script, and Create fills each beat with matching footage from your library, then lays it on the timeline.

PRO

Auto-Reel highlights

Pick a long recording and Create ranks its best moments as short highlights you can drop onto the timeline.

PRO
🎬

Editor / NLE export

Send search results or a Create timeline to Premiere, Resolve or Final Cut as a frame-accurate FCPXML or EDL timeline.

Underneath all four

One library, kept in order

Organize what you've indexed, see how it connects, clear out duplicates and bring in public media — then let your AI tools search it too.

PRO
📚

Collections

Group files into your own collections and nest them like folders. Categories, tags and saved searches stay free.

PRO
🕸️

Knowledge map

See how your library connects — files linked by shared topics, speakers and people, or by meaning — beside a dashboard of what it holds and a curation inbox.

PRO
🧹

Smart cleanup

Exact duplicates are found for free. Pro adds near-duplicates, similar videos and blurry or low-quality files, with keeper suggestions — and every candidate goes to a reversible quarantine before anything is deleted.

PRO
⬇️

Media downloader

Pull public media from a link so it's transcribed and searchable like everything else.

🤖

Works with AI agents (MCP)

Connect Claude, Claude Code or Cursor in Settings → Connect to AI agents, and they can search, ask and read transcripts on your computer. Your files stay put; the assistant gets the results it asks for. Set it up →

🗄️

Back up, move, relink

Export your library as a portable ZIP — the originals stay where they are — import it on another computer, and relink media you moved to a new folder.

The AI runs here, not there

Local AI, zero cloud

Everything that makes MediaFind smart — transcription, semantic search, summaries and the Ask assistant — executes on your own machine. Download the models once, then unplug the internet and it still works.

  • Local language model. Ask and summaries run on an on-device LLM — a small one ships with the app, and larger ones are a one-time download — with no API keys, no per-token billing, and no prompt ever leaving your computer.
  • Transcription on your machine. Whisper models with word-level timestamps; on Apple Silicon, the whisper.cpp engine can run them on the GPU.
  • Choose your engine. Pick a transcription and assistant model that fits your machine on first run — speed or accuracy, your call.
  • Verifiable. The core path opens zero external connections — and a built-in privacy check lets you confirm it yourself, any time.
Ask your library● On-device · offline
What did we decide about the launch date?
The team agreed to push the launch to March 14 to allow a final QA pass before release.
retro.mp4 · 12:04standup_03.m4a · 04:31
🔒 Answered by a model running on your Mac — nothing left the device.
Who it's for

Built for anyone with hours of media to dig through

If you've ever scrubbed through a recording trying to find the one part you remember, MediaFind is for you.

🎙️

Podcasters & creators

Resurface that perfect quote across hundreds of episodes, then clip it for promos — without re-listening to anything.

🎬

Video editors

Find b-roll and soundbites by describing what's on screen, then export the selects as an FCPXML/EDL timeline to Premiere or Resolve.

📰

Journalists & researchers

Search hours of interviews by meaning, find every mention of a topic, and cite the exact timestamp it was said.

⚖️

Legal & compliance

Review depositions and recordings privately, on-device, where the material can never leave the machine.

🏢

Teams & meetings

Transcribe meetings live — no bot joins, and in the Mac app it hears both sides of the call — then turn them into action items and decisions, and ask across every meeting you've ever recorded.

📚

Lifelong archivists

Make a lifetime of home videos, photos, lectures and voice memos finally searchable — by what happened, not by date.

Local-first, and checkable

What leaves your machine, and when

Not just a claim — it's verifiable at runtime. The core path uses no account and no API key, and sends nothing; anonymous usage statistics stay off unless you turn them on.

What you doWhat leaves your machine
Core: index, search, ask, summarize, export, clip, notes & conversationsNothing
Subscription activation and update checksSubscription key/status or the current app version, only when requested
Anonymous usage statistics (off by default)Only if you turn them on: daily counts, never your files, searches, or any ID
Download a public link you paste inOnly that public URL — keyless, opt-in
Connect an AI assistant over MCP (opt-in)The results it asks for go to that assistant — and on to its provider, if it's a cloud AI
People & FacesNothing — on-device, toggle off anytime
First run of optional ML & language modelsOne-time public model download, then fully offline
Live privacy checkRun the core features and MediaFind confirms it: 0 external connections.
Verified on-device
Why MediaFind

The privacy of local, the power of the cloud

Cloud media tools upload your files to someone else's servers. MediaFind keeps the media and AI processing on your machine — and unlike other local apps, you can verify it.

MediaFind Cloud media tools
Your files stay on your machineYesUploaded
AI assistant runs locallyYesServer-side
Works fully offlineYesNo
Account & API keys requiredNoneRequired
PricingMonthly or annualUsually monthly
Per-minute processing feesNoneUsually metered
Export to FCPXML / EDL timelineYes (Pro)Rare

See how MediaFind compares to Jumper & Clipto →

Pricing

Try Pro free. Keep it monthly or save annually.

Fresh installs include a separate 7-day no-card local Pro trial. Starting either Pro plan begins its own 7-day Stripe billing trial before the first charge. Cancel anytime.

Free

Try all four parts on a real library.

$0
  • Index up to 10 media files — video, audio or photos
  • Memory: chats and notes, which don't count toward the 10
  • Search by words, visuals, on-screen text and speakers
  • On-device Ask, summaries & chapters
  • Create's timeline editor, video tools & exact-duplicate cleanup
  • Keeps working after a Pro trial ends · no account
Download free
First month $9.99

Pro Monthly

Maximum flexibility after a free 7-day trial.

$9.99 first month
  • Unlimited indexing and every Pro feature below
  • $24.99/month after the first month
  • Cancel anytime
Start monthly trial
64% off

Pro Annual

The best value after a free 7-day trial.

$8.99 /month
  • Unlimited indexing and every Pro feature below
  • Billed $107.88 yearly
  • One simple annual renewal
  • Cancel anytime
Start annual trial

Everything in Pro

Both plans unlock all of it, and all of it runs on your computer.

  • Unlimited indexing — no 10-file cap
  • People & Faces — every appearance of a person, public figures auto-named
  • Meetings mode — live notes, action items and decisions
  • Brand-logo detection — by name or by a sample logo
  • Media downloader — public media from a link
  • Collections — your own groupings, nested like folders
  • Knowledge map — how your library connects
  • Highlights & reels — best moments stitched into one video
  • Smart cleanup — near-duplicate and quality culling
  • Editor / NLE export — FCPXML and EDL for Premiere, Resolve & Final Cut
  • Compose storyboard — a first cut from an idea or a script
  • Auto-Reel highlights — short highlights from a long recording
FAQ

Questions, answered

Everything you might want to know before you download.

Does anything get uploaded to the cloud?
No. Indexing, transcription, AI understanding, search and Ask all run on your computer — including the local language model that powers Ask and summaries. The only network activity is opt-in and predictable: a one-time download of optional ML models on first run, license activation and update checks, anonymous usage statistics if you turn them on, and any public URL you explicitly ask MediaFind to fetch. If you connect an AI assistant over MCP, the results it asks for go to that assistant. A built-in privacy check confirms zero external connections on the core path.
Does the AI run locally too?
Yes. Transcription runs Whisper models on your computer — on Apple Silicon you can switch to the whisper.cpp engine, which uses the GPU — and Ask and summaries use an on-device language model: a small one ships with the app, and larger ones are a one-time download. No prompts, transcripts or media are sent to a server.
What is Memory?
Memory keeps your AI conversations and notes next to your media. Import a ChatGPT or Claude data export, a text file or a pasted chat, or a folder of Markdown notes such as an Obsidian vault, then search and ask them: an answer lists the messages it read, or says your conversations don't cover the question. Export a library, delete a conversation or reset a whole library whenever you like. It all stays on your computer, it's free, and chats and notes don't count toward the free plan's 10 files.
What can I make with Create?
Create is the timeline editor built into MediaFind. Search your library from the side panel, drop clips on a multi-track timeline, cut by editing the transcript, add captions and titles, reframe for vertical or square, and render MP4 or ProRes. With Pro, describe a video or paste a script and Create assembles a first cut from your footage, Auto-Reel picks highlights from a long recording, and the timeline exports as FCPXML or EDL for Premiere Pro, DaVinci Resolve or Final Cut Pro.
Can I export moments to my video editor?
Yes, with Pro. Select any set of search results and export them as an FCPXML or EDL timeline that opens directly in Premiere Pro, DaVinci Resolve, or Final Cut Pro — with frame-accurate in/out points, snapped per each file's frame rate.
What platforms does MediaFind support?
MediaFind is a native macOS app for Apple Silicon, signed and notarized, for macOS 14 (Sonoma) or later. Windows and Linux builds are in beta and, like the Mac app, include ffmpeg for clip export and the video tools, Tesseract for on-screen text search, and phone access. The mobile companion, also in beta, lets you search your library and read transcripts from your phone over Wi-Fi, connected to MediaFind on your computer. Android installs from an APK; iOS is planned for TestFlight.
How does the Pro subscription work?
Choose monthly at $24.99 — $9.99 for the first month — or annual at $107.88. Both plans include a 7-day free trial and unlock unlimited indexing plus every Pro feature while your subscription is active.
Is there a free trial?
Yes. Fresh installs include a separate 7-day no-card local Pro trial. Starting a subscription begins its own 7-day Stripe billing trial before the first charge. The monthly plan is $9.99 for the first paid month and $24.99 per month after that; the annual plan is $107.88 per year. Cancel before the billing trial ends to avoid a charge.
What files can I index?
Video (MP4, MOV, M4V, MKV, WebM), audio (MP3, M4A, WAV, AIFF, FLAC, OGG) and photos (JPG, PNG, HEIC, WebP, GIF, TIFF, BMP). Point MediaFind at a folder and it indexes what it finds. While the app is open it keeps watching the folders you've added: it checks them about once a minute and indexes a new or changed file once the file has finished copying and any indexing already under way is done. You can turn this off in Settings. Chats and notes come in through Memory.
How does face recognition work — and can I turn it off?
People & Faces is a Pro feature that runs entirely on your device — nothing about your faces ever leaves your machine. For Pro libraries it runs automatically, and you can switch it off or delete all stored face and people data at any time with a single click. Free libraries are never face-indexed.
How do refunds work?
Payments, subscription changes and eligible refunds are handled through Stripe. Reach us at the support email from the address used at checkout and we'll help.
Get started

Make your media library searchable today

Free trial, no account required. The Mac app is signed and notarized; the Windows and Linux apps are in beta.

Apple Silicon (macOS 14 or later) · Windows x64 · Linux x64. Windows, Linux and Android builds are unsigned and may show an install warning.
The mobile companion searches your library and shows transcripts on your phone over Wi-Fi. It connects to MediaFind on your computer (Settings → Phone access). iOS is planned for TestFlight.