v1.32.2

• Remove support for CogVLM2 on Windows (#341)

v1.32

• Add support for CogVLM2 on Windows (#326)
• Use safetensors versions of models if available (#174)

Windows

Download taggui-v1.32.2-windows.7z and extract the files. You may have to install 7-Zip if it is not installed.

Linux

The bundle is split into two files to stay under GitHub's 2 GB file size limit. Download both parts (taggui-v1.32.2-linux.7z.001 and taggui-v1.32.2-linux.7z.002) and extract the files using the following command:

7z x taggui-v1.32.2-linux.7z.001

You may have to install 7-Zip if it is not installed.

v1.32.1

• Attempt to fix error with Triton (#341)

v1.32

• Add support for CogVLM2 on Windows (#326)
• Use safetensors versions of models if available (#174)

Windows

Download taggui-v1.32.1-windows.7z and extract the files. You may have to install 7-Zip if it is not installed.

Linux

The bundle is split into two files to stay under GitHub's 2 GB file size limit. Download both parts (taggui-v1.32.1-linux.7z.001 and taggui-v1.32.1-linux.7z.002) and extract the files using the following command:

7z x taggui-v1.32.1-linux.7z.001

You may have to install 7-Zip if it is not installed.

v1.32.0

• Add support for CogVLM2 on Windows (#326)
• Use safetensors versions of models if available (#174)

Windows

Download taggui-v1.32.0-windows.7z and extract the files. You may have to install 7-Zip if it is not installed.

Linux

The bundle is split into two files to stay under GitHub's 2 GB file size limit. Download both parts (taggui-v1.32.0-linux.7z.001 and taggui-v1.32.0-linux.7z.002) and extract the files using the following command:

7z x taggui-v1.32.0-linux.7z.001

You may have to install 7-Zip if it is not installed.

v1.31.0

• Add support for the JoyCaption auto-captioning model (#263, #266, #287) (#311 @maedtb)
• Allow using newline as a tag separator (#244, #267, #305)
• Restore functionality of LLaVA v1.6 models (#303)
• Update the moondream2 auto-captioning model to the latest revision
• Improve experience when running multiple instances of TagGUI (#308 @maedtb)
• Fix WD Tagger image preprocessing bug (#310 @maedtb)

Windows

Download taggui-v1.31.0-windows.7z and extract the files. You may have to install 7-Zip if it is not installed.

Linux

The bundle is split into two files to stay under GitHub's 2 GB file size limit. Download both parts (taggui-v1.31.0-linux.7z.001 and taggui-v1.31.0-linux.7z.002) and extract the files using the following command:

7z x taggui-v1.31.0-linux.7z.001

You may have to install 7-Zip if it is not installed.

v1.30.1

• Add SmilingWolf/wd-eva02-large-tagger-v3 to models list (#257 @yggdrasil75)
• Update packages to newer versions

Windows

Download taggui-v1.30.1-windows.7z and extract the files. You may have to install 7-Zip if it is not installed.

Linux

The bundle is split into two files to stay under GitHub's 2 GB file size limit. Download both parts (taggui-v1.30.1-linux.7z.001 and taggui-v1.30.1-linux.7z.002) and extract the files using the following command:

7z x taggui-v1.30.1-linux.7z.001

You may have to install 7-Zip if it is not installed.

v1.30.0

• Add support for more auto-captioning models
  • Phi-3-Vision-128K-Instruct (#158, #246)
  • MiaoshouAI/Florence-2-base-PromptGen (#243)
    • To generate tags with this model, use the default prompt (leave it empty) and uncheck Remove tag separators in caption
  • SmilingWolf/wd-vit-large-tagger-v3
• Add F5 shortcut for reloading directory in addition to Ctrl+Shift+L (#221 @yggdrasil75)

The bundle is compressed as a .7z file to stay under GitHub's 2 GB file size limit. You may need to install 7-Zip on your system to extract the files. Download the file for your operating system, extract it, then run the executable file inside.

v1.29.0

• Add support for Florence-2 auto-captioning model (#200)
• Add internlm/internlm-xcomposer2-vl-1_8b to the models list (#208)

The bundle is compressed as a .7z file to stay under GitHub's 2 GB file size limit. You may need to install 7-Zip on your system to extract the files. Download the file for your operating system, extract it, then run the executable file inside.

v1.28.0

• Add option to use regex for find and replace (#191 @yggdrasil75)
• Add option to show an alert when auto-captioning is finished (#172 #180 @geroldmeisinger)
• Show start time, end time, and average time per image when auto-captioning multiple images (#179 @geroldmeisinger)
• Reduce directory loading time (#162 @yggdrasil75)
• Fix errors caused by empty strings in "Discourage from caption" and "Include in caption" (#194)
• Handle errors when getting Exif tags (#175 #178 @geroldmeisinger)

The bundle is compressed as a .7z file to stay under GitHub's 2 GB file size limit. You may need to install 7-Zip on your system to extract the files. Download the file for your operating system, extract it, then run the executable file inside.

v1.27.0

• Add support for the InternLM-XComposer2-4KHD auto-captioning model (#120, #131)
  • Unquantized only; requires 24+ GB of VRAM
• Fix Triton error for CogVLM2 in the Linux bundle (#177)

The bundle is compressed as a .7z file to stay under GitHub's 2 GB file size limit. You may need to install 7-Zip on your system to extract the files. Download the file for your operating system, extract it, then run the executable file inside.

v1.26.0

• Add support for the CogVLM2 auto-captioning model (Linux only) (#149)
• Allow using the backspace key to delete tags (#155 @Cernunnos)
• Fix outputs for the LLaVA-NeXT (LLaVA v1.6) Vicuna auto-captioning models

The bundle is compressed as a .7z file to stay under GitHub's 2 GB file size limit. You may need to install 7-Zip on your system to extract the files. Download the file for your operating system, extract it, then run the executable file inside.