Keep your AI generations output organized — on your own PC, only one click away.

I was constantly searching for things and sometimes didn’t find what I was looking for. Media management was a huge time sink, and I’d rather spend that time generating than searching for that one specific file. genvault helps with that.

genvault is a small local web app for Windows that catalogs your AI-generated videos, images and sounds. It reads the generation data embedded in your files (prompt, model, seed, sampler, LoRAs …), makes everything searchable and lets you sort your work into projects. Your files stay on your disk and are never uploaded anywhere.

Current version: 2.2.0 · Windows 10/11 · MIT License · vibe coded with Claude

Features

Organize & Manage

• Drag & drop files into the browser window, or drop files and whole folders into a watched drop folder

• Everything lands in an Inbox first; you file it into Project / Sub-project / Kind yourself, so nothing gets sorted wrongly

• Bulk filing, moving and deleting for many files at once; files are physically organized in matching folders

• Rename a project at any time; its files are moved along (renaming onto an existing project merges the two)

Generation metadata

• Reads prompts, negative prompts, model, seed, sampler, scheduler, steps, CFG, shift and LoRAs from ComfyUI videos and PNGs (including Kijai WanVideoWrapper workflows) and A1111 PNGs

• ⚠️ Metadata reading is best effort and won't work for every workflow (see the disclaimer under Good to know). genvault is a media organizer, not a dedicated prompt manager.

• Every field is editable, plus a notes field for things like Suno styles

• Copy prompt, Copy all parameters and Copy workflow JSON buttons

Find anything again

• Full-text search across prompts, tags, projects, LoRAs and file names

• Filter sidebar with counts: project, sub-project, kind, media type, character, LoRA, upload status

• Star ratings, favorites, "Top rated" sort

• Video previews on hover, audio waveforms with a play button

Release & Share

• Release module: add your PNG watermark logo (9 positions, size, opacity, margins, live preview frame), join several clips into one video, render in the background on your NVIDIA GPU (NVENC) with automatic CPU fallback

• Lossless joining when clips match; three quality presets otherwise

• Embedded prompts and workflows are stripped from rendered files, so you never leak your workflow by accident

• Track where each release was posted (X, Civitai, Reddit by default) with URL and date

Peace of mind

• Privacy blur: one click blurs every preview (handy when screen sharing)

• Deleting moves files to a Trash, where you can restore or permanently delete them (one by one or many at once)

• One-click metadata backup; the database can always be rebuilt from per-file JSON sidecars

• Safe updates: a new version replaces only program files, never your library

Requirements

• Windows 10 / 11

• Python 3.10+ ([python.org](https://www.python.org/downloads/)), tick "Add python.exe to PATH" during installation

• ffmpeg ([gyan.dev builds](https://www.gyan.dev/ffmpeg/builds/), "release essentials"): copy ffmpeg.exe and ffprobe.exe into C:\ffmpeg\bin

• Optional: an NVIDIA GPU for fast release rendering

• An internet connection when opening the app (the page styling is loaded from a CDN); your media never leaves your PC

Installation

1. Download and unzip. Move the genvault folder somewhere permanent (e.g. D:\genvault).

2. Double-click setup.bat. It sets up its own Python environment and asks where your library should live (pick a drive with space) and which port to use (just press Enter).

3. Double-click genvault-tray.bat. When the tray icon turns green, right-click → Open genvault (or go to http://127.0.0.1:8077).

Full instructions, configuration and troubleshooting are in the included README.md.

Updating

Quit genvault from the tray icon, drag the new zip onto update.bat, start again. Your library, settings and database are kept.

Good to know

• genvault runs only on your machine 127.0.0.1). There are no accounts and no telemetry.

• Files are filed manually by design; genvault never guesses projects from file names or prompts.

• No duplicate detection and no settings screen: platforms and kinds are set in config.json.

⚠️ Disclaimer: prompt and metadata detection

There are countless ComfyUI nodes and custom workflows, and there is no standard for where a prompt is stored. genvault recognizes common setups, but it cannot reliably read every workflow. Depending on your nodes you may see:

• Correct values: typical for A1111 images and standard ComfyUI workflows

• Empty fields: the file has no embedded workflow (e.g. saved without metadata, or re-encoded/edited afterwards), or the prompt comes from a node genvault doesn't know

• The negative prompt in the positive field (or vice versa): genvault can't always tell which text node is which, especially when both still have their default name

Tip: In ComfyUI, rename your text-encode nodes so the titles contain "Positive" and "Negative". genvault uses these titles first, which makes detection much more reliable.

Wrong or missing values can always be corrected by hand in the details panel. When a workflow is embedded, Copy workflow JSON still gives you the complete original data. Editing metadata in genvault never modifies the contents of your files.

What's new in 2.2.0

- Rename projects right from the sidebar (pencil icon on hover)

- Multi-select delete: move many files to the Trash at once, and restore or permanently delete many files in the Trash

- The filter sidebar keeps its scroll position instead of jumping back to the top

Fixes since 2.1

- Filing, moving or deleting a file while its preview is playing no longer fails with "file is used by another process"

- No more HTTP 500 errors on delete/restore of locked files

- No duplicate files after a failed move; database and folders always stay in sync

- The autocomplete list in the bulk bar now docks to the bar correctly

- Clearer error messages that name the files that failed

Feedback and bug reports are welcome. Please include the version number shown next to the logo.

Released under the MIT License. Free to use, modify and share.