Skip to content

Releases

Media Manglers v0.11.0 (beta)

Media Manglers v0.11.0 (beta) Pre-release
Pre-release

Choose a tag to compare

View all tags
@SteveSobka SteveSobka released this 10 Nov 00:23
v0.11.0
6051158
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

This is the first public download of Media Manglers. It is a beta for 64-bit Windows 10 and 11: it works on real recordings, and it is published so more people can run it and it can be improved from what they find.

What it does

The Video Mangler Workbench is one app with four features on one shared engine. Everything runs locally on your PC. Your recordings are only read, never changed, and nothing leaves your computer unless you turn on an AI mode.

  • Package any video for AI. Any video, a file or a web link, racing or not: screenshots, a transcript and clips in a package an AI can review, even when the video itself is too big to upload.
  • Find my bug reports. Narrate bugs while you use your software, press one button, and paste a ready prompt into Claude Code or Codex. Your agent reads the package on your PC, drafts the issues, and opens them only after you approve.
  • Watch the raw bin. A library of every recording in your recordings folders: what is in each one, how much is idle, and what you decided to keep. The Review Media Inbox reviews new recordings and remembers which are done. The app only marks; it never moves or deletes a recording.
  • AI Video Editor. Turn a long race recording into a story-driven video ready for YouTube. Your AI agent edits; you watch each version, say what to change, and export the one you like. Your private camera areas are always blurred.

Before you start

The download holds the Workbench itself. To review recordings it also needs these tools, which you install yourself:

  • FFmpeg (it includes FFprobe), to read video and audio.
  • Python 3.12 with Whisper and PyTorch installed into it with pip, to write transcripts on your PC. The Whisper model is downloaded only after you approve it in Settings > Dependencies.
  • yt-dlp, only if you want to review web-video links.
  • VLC, optional, only for Play with > VLC.

It also uses Windows PowerShell, which is built into Windows.

After the first start, open Settings > Dependencies. It shows what is missing or out of date. Download opens each tool's official page, and Copy winget command copies the install command where Windows Package Manager has the tool. The app never downloads or installs a tool by itself. A one-step installer that sets these up for you is planned.

How to install

  1. Under Assets below, download VideoManglerWorkbench-v0.11.0-windows-x64.zip.
  2. Right-click the ZIP, choose Extract All..., and extract all of it to a short folder such as C:\MediaManglers. Do not run the app from inside the ZIP: it needs every file and folder that sits next to it.
  3. Open the extracted VideoManglerWorkbench-v0.11.0 folder and run VideoManglerWorkbench.exe.
  4. If Windows shows "Windows protected your PC", choose More info, then Run anyway. This appears because the app is not code-signed yet.

To check that your download is complete and unchanged, open PowerShell in the folder that holds the ZIP and run:

Get-FileHash .\VideoManglerWorkbench-v0.11.0-windows-x64.zip -Algorithm SHA256

The hash it prints must match the checksum at the end of these notes, which is also in SHA256SUMS.txt. Upper or lower case does not matter.

What's new

v0.11.0 follows v0.9.6 beta 1. v0.10.0 was prepared but never published, so its changes are in this release too. The highlights:

  • A new look for every screen, with a Text size setting (100%, 110%, 125% or 150%) that changes the whole app at once.
  • The raw bin library: one lasting list of every recording in your recordings folders, with the app's guess, idle time, your decision and labels.
  • The AI Video Editor's own page, built around its player: versions as tabs, the story as a strip of scenes, privacy masks that the AI editor cannot remove, and Approve and export... as your approval.
  • Waveforms of every audio track in the built-in player.
  • A History that keeps every run and says what really happened to each recording.
  • Find my bug reports needs no microphone step, keeps its lists in the library, and is about a quarter faster.
  • Words to spell right: one list of your own names and terms that every feature gives to Whisper. Lines Whisper invents are left out, each listed with its reason.
  • A clear message when a tool is missing, with Download and Copy winget command.

Every feature is described in one guide, the feature guide, which also opens from Help in the app.

If you used an earlier build: the first start of v0.11.0 moves your run history and Find my bug reports' lists into the app's library, after making a checked backup. Read "Read this before you upgrade" in the full release notes first, and do not go back to an older build afterwards without following its restore steps.

Full release notes: RELEASE_NOTES_v0.11.0.txt

Known limits

  • A Raw bin scan reports progress only when a whole recording is done, so a very large recording can look frozen for minutes while it works. After Cancel review, its count of finished items is wrong.
  • If the library cannot be opened when the Workbench starts, the Bugs page shows its never-use and already-done lists as empty for that session, and nothing on screen says why. Nothing is lost: they come back at the next start that opens the library.
  • AI Video Editor: in some recordings the sound runs up to about one and a half seconds ahead of the picture, and the editor does not correct it yet. A replay or a moving menu looks like driving, so dead time there is not found. Its runs can be cancelled but not paused.
  • Raw bin: Deep scan does not run the cross-track check and does not give Whisper your words to spell right. Idle time comes from the game sound only. Delete candidate is only a mark; deleting marked files is not built.
  • Review Media Inbox keeps its own record of done recordings, apart from the library, for now.
  • Next sound finds sound, not speech. On a microphone that also hears the game or a fan it can land a few seconds off.
  • A full mix with no speech can still get invented lines.
  • Find my bug reports keeps its package as a folder; an optional ZIP export is not built yet.
  • For people running the test suite: a full parallel run takes about two hours instead of under one. The app itself is not affected.

Licence

  • Code, documentation text and screenshots: GNU AGPL-3.0-or-later. The licence text ships in the download as LICENSE.txt.
  • Third-party parts the package bundles (Qt for Python, the FFmpeg libraries inside Qt, the Python runtime and others) keep their own licences. See THIRD_PARTY_NOTICES.txt and the licenses folder in the download.
  • Names and icons: the AÏntologic name and logos, the Media Manglers, Video Mangler, Audio Mangler and Video Mangler Workbench names, and the app icons are not covered by the AGPL. You may share unmodified copies of the app with its icons; see TRADEMARKS.md for the rest.

Feedback

Report bugs and request features on GitHub Issues. Include the version shown in the bottom-left of the Workbench. Please do not post private recordings, credentials or confidential output.

Checksum

SHA-256 of VideoManglerWorkbench-v0.11.0-windows-x64.zip:

d3e1dd3c17f669532b4ef9c0c933bc3076e845334665588b2135e27c55721a88
Loading