Skip to content
soybean-alienPublic

About

Turn any video or picture into a titled, timestamped note on your own Windows PC. .NET 10, WinUI 3, local Whisper. GPL-3.0.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

5 Commits

Folders and files

Repository files navigation

Mimic

Mimic

Built with C# 93.5%, PowerShell 3.5%, NSIS 1.7%, HTML 1.3% (GitHub language breakdown)

Turn any video or picture into a titled, timestamped note, on your own Windows PC, with no subscription.

Windows .NET 10 and WinUI 3 License: GPL-3.0

Mimic is a small Windows desktop app built with .NET 10, WinUI 3 and the Windows App SDK. Paste a link, or point a crosshair at a video that has no link, and Whisper writes the transcript on your own machine. A model of your choice then titles every section, and one click saves it all as Markdown for your vault or note program.

Install

  1. Download Mimic_0.3.2_x64-setup.exe from the latest release. It is about 7 MB.
  2. Run it. Windows may show a "Windows protected your PC" box because the installer is not signed; choose More info, then Run anyway.
  3. Open Mimic and click Download tools, then Download Whisper model.

New here? Read START HERE.md, a step-by-step guide in plain English.

What the installer contains, and what it downloads

The installer contains only Mimic itself. Everything else is downloaded when it is needed:

Item Size (estimated) When it is downloaded
.NET 10 Runtime, x64 (Microsoft) 30 MB Only if this PC does not have it: setup opens Microsoft's download page for it in your browser and waits while you install it
Windows App Runtime 2.6 (Microsoft) 125 MB Only if this PC does not have it: setup opens Microsoft's download in your browser and waits while you install it
yt-dlp 23 MB In Mimic, the first time it is needed
FFmpeg and ffprobe 193 MB In Mimic, the first time it is needed
Whisper engine (whisper.cpp) 10 MB In Mimic, the first time it is needed
Whisper model (ggml-large-v3-turbo-q5_0.bin) 574 MB Optional. Only when you choose to download it in Mimic

The Whisper model is not in the installer. Mimic needs it to transcribe, so the first time you transcribe (or when you open Downloads from the missing-tools notice at startup) Mimic offers to download it. The download is checked against its SHA-1 before it is used. yt-dlp, FFmpeg and the Whisper engine are each checked against the SHA-256 that GitHub publishes for that release file; a download that does not match is deleted, and if GitHub gives no checksum nothing is downloaded. If the model is already in %USERPROFILE%\Mimic\models, Mimic uses it and downloads nothing.

The simple way is two clicks in Mimic's Downloads box: Download tools, then Download Whisper model. To fetch the files yourself instead, the Downloads box links to each one. The two single-file downloads are:

Reading, exporting and playing transcripts you already have works without any of the downloads. Tools already in %USERPROFILE%\Mimic\tools are used as they are, including a GPU Whisper engine if you put one there; the engine Mimic downloads itself is the CPU build.

Features

  • Transcribe from a link. yt-dlp fetches the audio and whisper.cpp transcribes it on your machine.
  • Embedded video extraction. Point a crosshair at a video on your screen, link or no link. Mimic tries available video sources, reports extraction attempts and results, and falls back to listening to the window when video extraction fails. Listening runs in real time.
  • Drop a file. Drop a video file on the window to transcribe it, or a picture to describe it.
  • Video length and host. Shows the video duration and sanitized hosting names when the source provides them.
  • Local video playback. Plays saved video downloads from the transcript; timestamps jump to the matching section.
  • Progress and timer. A progress bar and timer run while video jobs are working.
  • Key takeaways by topic. Timestamped topic sections with supporting details.
  • Section titles. Every timestamp gets a short title saying what that section covers.
  • Save .md. A note with frontmatter, a heading per section, and timestamps that link back to that second on YouTube. Save .md opens Save As, so you choose where the note goes. Optional direct filing: if the MIMIC_OBSIDIAN_VAULT variable is set when Mimic starts, the note goes straight into that folder's Claude Code/Development/Projects/Mimic/Projects subfolder instead, which must already exist; if the vault contains .obsidian, Mimic asks before writing.
  • More exports. Copy, Save .srt, Save .txt and Export PDF.
  • Pictures. Browse to an image for a very detailed reverse prompt, plus an AI score in one of six tiers of certainty. The score is a model's opinion, not a measured detector.
  • Two tabs. Run two jobs side by side, each with an MM:SS stopwatch.
  • A local library. Everything you make stays in a SQLite file on your PC.
  • Open Stored files. Opens Mimic's local video downloads folder in Explorer.
  • Change library folder. Pick your own folder for the library (library.db and pictures). Mimic copies the current library there if the folder has none, then restarts. Tools and models always stay in %USERPROFILE%\Mimic.
  • Update yt-dlp. One button in the sidebar.

Choose your model

Whisper always does the transcribing. The Model button picks what writes the titles, reverse prompts and score:

Choice Where your text and images go What it needs
Default: Haiku for titles, Sonnet for images Anthropic The Claude Code command
One Claude model for everything Anthropic The Claude Code command
OpenAI OpenAI The Codex command, signed in with your ChatGPT account; no API key
DeepSeek DeepSeek An API key
Ollama, LM Studio or another address Your own server A running server and model

Those sign-ins, keys and servers are not bundled. A new install starts on Default, signed out of everything.

Build from source

You only need this to change Mimic; to use it, download the installer above.

You need the .NET 10 SDK. Build output goes to %USERPROFILE%\Mimic\build\winui, outside the source tree.

dotnet build Mimic/Mimic.csproj -c Release -p:Platform=x64
dotnet test Mimic.Tests/Mimic.Tests.csproj

To pack the installer you also need NSIS 3. The script finds makensis.exe in NSIS's default install folder or on PATH; for any other location pass -MakeNsis <path>:

powershell -ExecutionPolicy Bypass -File installer/build.ps1

The installer is written to %USERPROFILE%\Mimic\build\winui\installer.

Good to know

  • Mimic keeps its data in %USERPROFILE%\Mimic: library.db, images, tools and models. Uninstalling removes only the program files; the library, tools, models and saved videos stay.
  • API keys for the address route are stored as plain text in library.db.
  • Whisper is unreliable on music and makes mistakes on poor audio.
  • Playing a video faster while Embedded listens compresses its timestamps.

Credits and AI disclosure

  • Made by soybean_alien. Licensed under the GNU General Public License v3.0 (see LICENSE).
  • .NET, Windows App SDK and WinUI 3, Microsoft.Data.Sqlite, yt-dlp, FFmpeg and whisper.cpp.
  • Mimic is an unofficial personal project and isn't affiliated with Anthropic, OpenAI, DeepSeek or Microsoft.
  • AI disclosure: Mimic was designed with and written by Claude Code (Anthropic, Claude Fable 5.2), directed and tested by the author. Run in the real installed app (0.3.0 and 0.3.1; in 0.3.2 link transcription, the video player and Delete were run again): link transcription, a dropped video file, a picture's reverse prompt and score, section titles, Key takeaways, Copy, Save .srt, Save .txt, Save .md, Export PDF, the video player, Embedded capture (both extraction from a browser tab and listening to a window), Delete, opening a library, the Model dialog, the Downloads box and changing the library folder. The AI features were run on the Default Claude route only. Not yet run in the real app: the OpenAI, DeepSeek and own-server routes, OpenAI sign-in, and a fresh install on a PC without the runtimes.

About

Turn any video or picture into a titled, timestamped note on your own Windows PC. .NET 10, WinUI 3, local Whisper. GPL-3.0.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages