---
title: Comfyui-PNG-WebP-Mp4-Prompt Extract-V2-PlusPNG2WebP-tool
emoji: 😻
colorFrom: green
colorTo: green
sdk: static
pinned: false
license: apache-2.0
short_description: Extract prompts and settings-NEW -still 100% local&private
---
Now V2. Much better internally at extracting prompts, even from multi prompt workflows. Also showing all prompts found. Improved UI.  STILL offline, private, using Local browser storage only!
--
like pls if you do. 

- Under the hood it went from guessing which big text block is which (V1, hit or miss) to analyzing the workflow data to find all the correct ones (related to clip text encoders).  
- With the rise of prompt enhancing nodes, LLM nodes, Zit Upscalers, multi step video workflows and so on, the old version became much more miss than hit.  There are a LOT of plain text bits in such a  workflow. 
- Also the UI is less messy and now displays all prompts in separate fields with some QOL features.
Should all be fairly self explanatory.

V2 is ComfyUI only.  I dropped support for InvokeAi generated images.  
BUT prompt extractor V1 is still available as space (check under my user)  which supports ComfyUI AND invokeAI generated images (and comfy videos). It does it's job ok for InvokeAI images and Comfy ones from simpler workflows.  But it will fail quite often with complex ones.

In files is a new PNG to WebP converter tool  -   Converting Comfyui PNG to WebP (10% size)  while maintaining metadata and workflow data within!   That means the WebP can STILL be dropped into comfyui to reload the embedded workflow from the PNG.  And the prompt and settings can still be extracted. 
--

It's a tiny python tool with a simple web interface.  
- Download and extract zip locally
- It will need a local Python installation on path to create the .venv on first start.  
- Starting with .bat (Win, tested) or .sh (Linux, Mac, untested!) file.
- It needs internet access first run, to download the requirements (Pillow, about 30MB). It wil also create the .venv if there isn't one.  Than it will launch (local host, port will auto switch if already used).
- Important caveat:   It was made for the purpose of converting ComfyUI images, not as general converter. it will skip any PNG that aren't!  It is NOT designed to maintain other unique or special Metadata.

After the first run,  it works 100% offline and local. NOTHING leaves your computer.
Hardware:  Being 100% private and offline,  It uses YOUR CPU,  supporting multiple CPU cores (set or leave 0 (all).  I converted about 16.000 PNG to WebP on a Ryzen 9 9950X3D in about 15 Minutes.  Your mileage WILL vary. It is a bit slower than my old one (in the V1 space) but there is a LOT more going on internally.

The main extractor tool remains a single, self contained html file that can also be downloaded and run local and offline (just dl index.html under files. 

But no matter if you run it from the Huggingface space or local, ALL processing happens in your local browser using your local machine.  NOTHING is shared or uploaded.
--
Both tools are again Apache2.0 (do with it, what you want but don't blame me).  A link (to the space) or a mention would be nice though. Drop a like if you do.

Both were created with ChatGPT5.6 on sol high.  
It did the full conversion thing in minutes... i remember my LONG trial and error session with 5.0 or so. 
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference