Image Manipulate is a AIO (all-in-one) tool designed to help manage training dataset by:

tagging using Ollama
extracting frames from video source material
managing content of training dataset
edit image content
browse content of safetensor files (allow to view metadata and tensors, including image preview if available, has to have same base file name)
view media metadata
media metadata (file embedded) prompt (tags) and workflow extraction
multiple views (that can be re-arranged), column based views allow re-arrangement of columns
live update views when files in working directory are added / delete from outside of application (modification of file content is still not reflected, refresh directory)
live filter (using filter ":missingtxt" will also show newly added files that don't have description files)

This is "portable" application. Download -> eXtract -> Use. Application stores all its settings/logs in current directory so it needs write access, otherwise settings will be not saved, but app will still works. Settings are compatible with newer versions of applications so either overwrite new files (if library files are not changed) or copy old settings file to new version location to preserve old settings.

Currently this tool is available only on Windows OS.

NOTE: this app can also be used to manage Your images.

Usage note:

All deletion operations only move files to trash bin. If trash bin is not available deletion operation fails and file is left as is. This is "safety" measurement, at least for now.
changes done to the working directory from outside of application (i.e. adding new files via file browser) are not auto reflected, use refresh working directory functionality (all unsaved changes to the tags are discarded, save before refresh)
main window working directory image list support multi item selection and most operations are being done on selected items
most widgets, and tools like image edit, have custom context menu with additional options (right mouse button click)
most widget do have tool tips, hover over them to see description
Image cut action is triggered by either Enter or double clicking (anywhere inside edit image window)
Image cut undo CTRL+Z (not exposed in Ui)
Cut images are not automatically saved use Save (CTRL+S) or Save ass (SHIFT+S)
Image Edit - CTRL+WHEEL -> go to next / previous image

Changelogs:

0.7.0.130

Correcting Comment column to reflect recent support for webp and jpg files
Correcting workflow export for webp and jpg

0.7.0.128

Adding tag support for webp and jpg
Removing unused Qt dll shaving off size of app

0.7.0.125

Adding view menu allowing to switch views easier
Correcting workflow detection in file metadata info
Adding column with tags counter
Update txt file column upon detecting external changes to file system
Adding ability to extract tags from embedded workflow. Currently only support PNGs files. Duplicated entries controlled by save duplicate option in main window.
- NOTE: approach used actually parses workflow and handles basic nodes, using custom, complex, nodes for positive prompt could cause for this feature to not work. The simplest prompt route the more likely it will work correctly.
bug fixing index

0.7.0.105

Supporting Ollama text annotation for video files using single, first, frame.
Displaying total frame count as tooltip in Duration column
Adding ability to ignore large metadata data, Metadata tab -> large (checkbox - disabled by default). This causes Ui slowdown on multi-selection, especially if media do contains embedded workflows (that could be very large textual data).
Resolving issue with folder watcher (sometimes newly added video files did not triggered correctly update)

0.7.0.94

adding support for video media
adding support for video preview
adding ability to export workflow from media
- WIP feature, not working always as these workflows could be stored under different keys, all media metadata accessible by metadata tab with option to copy workflow if present and not correctly detected
adding multiple views to display data in different form / views
sorting enabled by columns using view headers (context menu, right mouse button, on header to access that menu)
ability to hide / show specific columns
column can be re-arranged / moved
media metadata tab that shows associated metadata to the selected media
rows editable so data can be copied or just CTRL+C while selecting row column to copy entire content (this apply to pretty much everything)
allows to copy, move selected files
editing txt file also have now separate widget same functionality like one on right side but as separate widget that allow easier focus on only text)

0.7.0.77

Main:

fixing issue with saving tag file. Sometimes, when tag file was missing, saving was done to previous, existing, txt file.
resolving issue with image preview loading in threads
adding ability to watch work directory, adding or removing files is reflected in Ui. NOTE: changes done to images and txt file content from outside of app (i.e. image changed in Krita / Gimp, txt file edited in notepad++) are NOT reflected in app.
image filter support custom tag ":missingtxt" that will show only images with missing tags txt files
windows can now be placed next to each other, vertical or horizontal.

Image Edit:

after saving file main window correctly update image data (size / thumbnail)

Image Effects (aka. filters):

adding support for filters stack
adding ability to move filter position on the stack
adding effects profiles

0.7.0.65

Main window
- image preview loaded in thread, exposing thread count in settings (work in progress)
- delete stray txt files counter correction

0.7.0.53

Main window
- Adding ability to cleanup stray txt files (txt files without image)
DSP Effects
- dsp effects available in main window tab. Currently only single effect is applied at the time in preview window (all when saving) as this is work-in-progress feature. Applying DSP effect in different order changes output and node approach will be used in the feature
Edit image
- menu entries that allow set same width / height for cut tool
- allows to keep cut rectangle aspect ratio (key A)
- toggling / using again Cut tool, via icon or keyboard shortcut, resets cut rectangle (type W) to its initial state
- exposing Undo functionality in Ui via dedicated button (and shortcuts CTRL+Z / platform Undo)
- triggering cut tool execution via mouse double click and keys like: space / enter / return.
Video widget
- replacing slider with progress bar for easier video navigation
- duration progress bar is now clickable, click and/or click + move changes media position
- keys 0-9 seek to % video duration, i.e. key 5 seeks to 50% of video duration
- Added tool "Grab series of frames" from currently open video file. Allows to specify number of frames to grab
- Added tool "Grab frames from files" s above, but allow to specify multiple video files

0.7.0.41

correcting "Open in native app" functionality.
Main window:
- Import files - exposed, in app options, remove source file, was on by default, when enabled source file is moved to trash, (if trash is not available file is left as is, this is global behaviour)
- Import files - exposed, in app options, auto rename destination file if file with same name already exist in working directory
Ollama:
- adding ability to abort queued prompt (there is delay from ollama response that is not handled, so don't spam prompt queue and look at window status message, on left bottom side of window, to know current status)
Image Edit:
- adding ability to move cut rectangle by arrow keys for more precise placement (arrow - 8px, CTRL + arrow increase movement speed to 40px, SHIFT + arrow forces 1px movement)
- adding ability to reset zoom (key R - globally available in all tools, menu for Pan tool -> right button context menu option)
- Cut/Pan tool - adding ability to restrict movement by X/Y plane. While moving cut rectangle / panning press SHIFT for X or CTRL for Y movement restriction.
- adding image option save as (default file name is based on original image file and is unique, format fileName_X.suffix), if saved in working directory file is added to image list

virustotal scan result 0.7.0.38.

virustotal scan result 0.7.0.41.

virustotal scan result 0.7.0.53.

virustotal scan result 0.7.0.65.

virustotal scan result 0.7.0.77.

virustotal scan result 0.7.0.94.

virustotal scan result 0.7.0.105.

virustotal scan result 0.7.0.125.

virustotal scan result 0.7.0.128.

virustotal scan result 0.7.0.130.

History URLs:

Files backup / historical versions / are exactly same as on civitai (prefer download from civitai) hosted on my server:

v_0.7.0.130

v_0.7.0.128

v_0.7.0.125

v_0.7.0.105

v_0.7.0.94

v_0.7.0.77

v_0.7.0.65

v_0.7.0.53

v_0.7.0.41

v_0.7.0.38

TODO's:

provide Linux (and, not likely but maybe, macOS) versions
improve image resizing by adding option of nearest aspect ratio scaling
add support external denoiser (i.e. Intel Open Image Denoiser)
move to multithreading for image loading (low priority, not needed as it is "fast" compared to training times
image DSP, i.e. levels, custom kernels support, colour gradient, etc.
...

Legal disclaimer:

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,

EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.

IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR

OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,

ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR

OTHER DEALINGS IN THE SOFTWARE.

Search AI Models & Addons

very few results