v4.5.1 (Flux Textual Inversion, minor fix)

v4.5.1 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.5.1 Patch

• Default --textual-inversions lexical token name for Flux should mimic behavior of the default token name for SDXL, i.e. the token name should be set to the model filename (spaces to underscores, no extension), when no token name is provided.

• Documentation fixes.

4.5.0 Features & Fixes

• Support for Flux fill (Infill / Outfill) via --model-type torch-flux-fill, this model type uses the inpainting --image-seeds syntax to specify an image and a mask. See: example

• Support for --textual-inversions with --model-type torch-flux-fill

• Support for --textual-inversions with --model-type torch-flux in basic txt2img configuration, with txt2img + Control Nets, and with inpainting + Control Nets.

• Support for a single --ip-adapters weights definition with --model-type torch-flux in basic txt2img configuration. Does not support Control Net or img2img / inpaint / fill. See: example

• Dynamic feature availability checking for LoRA, Textual Inversion, and IP Adapter support on supported diffusers pipelines using introspection, refactored input validation for these features in particular.

• Fix for unusable StableDiffusionControlNetPAGInpaintPipeline caused by typo.

• New Console UI recipes for --model-type torch-flux-fill

• #9, Remove problematic dependency bitsandbytes for now, dependency unused, and PyPI release for bitsandbytes currently incompatible with MPS and ROCm

• Update to diffusers==0.32.2, and update other pinned deps where possible.

• #8, Fix PyPI package python spec and related documentation, not compatible with python 3.13 due to dependencies, spec is now >=3.10,<3.13

• #7, recommend / use --extra-index-url https://download.pytorch.org/whl/rocm6.2/, latest ROCm backend for ROCm install

v4.5.0 (Flux Fill, Flux IP Adapter, Flux Textual Inversion)

v4.5.0 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.5.0 Features & Fixes

• Support for Flux fill (Infill / Outfill) via --model-type torch-flux-fill, this model type uses the inpainting --image-seeds syntax to specify an image and a mask. See: example

• Support for --textual-inversions with --model-type torch-flux-fill

• Support for --textual-inversions with --model-type torch-flux in basic txt2img configuration, with txt2img + Control Nets, and with inpainting + Control Nets.

• Support for a single --ip-adapters weights definition with --model-type torch-flux in basic txt2img configuration. Does not support Control Net or img2img / inpaint / fill. See: example

• Dynamic feature availability checking for LoRA, Textual Inversion, and IP Adapter support on supported diffusers pipelines using introspection, refactored input validation for these features in particular.

• Fix for unusable StableDiffusionControlNetPAGInpaintPipeline caused by typo.

• New Console UI recipes for --model-type torch-flux-fill

• #9, Remove problematic dependency bitsandbytes for now, dependency unused, and PyPI release for bitsandbytes currently incompatible with MPS and ROCm

• Update to diffusers==0.32.2, and update other pinned deps where possible.

• #8, Fix PyPI package python spec and related documentation, not compatible with python 3.13 due to dependencies, spec is now >=3.10,<3.13

• #7, recommend / use --extra-index-url https://download.pytorch.org/whl/rocm6.2/, latest ROCm backend for ROCm install

v4.4.2 (Dependency updates, URL Escaping & Web cache fixes)

v4.4.2 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.4.2 Patch

Update to latest diffusers release, update all dependencies where possible.

Ability to escape % and $ with \ to prevent environmental variable expansion in dgenerate config / shell language, this is needed to be able to utilize (download from) URLs that contain these characters. This escape mechanism is also understood from the command line if needed.

pygments (syntax highlighting) lexer updated to understand new escape sequences in URLs.

Minor fixes to on disk file cache behavior for downloaded files. Mainly issues arising from files
that are unexpectedly missing on disk but present in DB, that prevented self healing in certain cases.

4.4.0 Features

• --scheduler, --sdxl-refiner-scheduler, and --s-cascade-decoder-scheduler can now handle being passed multiple URIs, this allows you to iterate over different schedulers (samplers) directly from the CLI with one command and no further scripting. These arguments now also have pluralized aliases: --schedulers, --sdxl-refiner-schedulers, and --s-cascade-decoder-schedulers, to help with potentially annoying typos. Using multiple schedulers contributes to generation steps combinatorically, in the same way as other diffusion parameter arguments such as --inference-steps and --guidance-scales etc.

• New Edit -> Insert Argument Console UI dialog, you can now select dgenerate arguments from a drop down and insert them into the config. This new dialog has a help button which will display the help text for the selected argument from dgenerates --help output in isolation, similar to the help window available in the image processor selection UI. This dialog is also available by right clicking the text input area in the Console UI and selecting Insert Argument.

v4.4.1 (Scheduler batching / iteration, UI improvements) Patch

v4.4.1 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.4.1 Patch

Insert Argument Console UI dialog arguments are now sorted alphabetically, buttons are disabled if nothing is selected, and initial selection is set to empty.

Fix --help text formatting for --output-overwrite.

4.4.0 Features

• --scheduler, --sdxl-refiner-scheduler, and --s-cascade-decoder-scheduler can now handle being passed multiple URIs, this allows you to iterate over different schedulers (samplers) directly from the CLI with one command and no further scripting. These arguments now also have pluralized aliases: --schedulers, --sdxl-refiner-schedulers, and --s-cascade-decoder-schedulers, to help with potentially annoying typos. Using multiple schedulers contributes to generation steps combinatorically, in the same way as other diffusion parameter arguments such as --inference-steps and --guidance-scales etc.

• New Edit -> Insert Argument Console UI dialog, you can now select dgenerate arguments from a drop down and insert them into the config. This new dialog has a help button which will display the help text for the selected argument from dgenerates --help output in isolation, similar to the help window available in the image processor selection UI. This dialog is also available by right clicking the text input area in the Console UI and selecting Insert Argument.

v4.4.0 (Scheduler batching / iteration, UI improvements)

v4.4.0 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.4.0 Features

• --scheduler, --sdxl-refiner-scheduler, and --s-cascade-decoder-scheduler can now handle being passed multiple URIs, this allows you to iterate over different schedulers (samplers) directly from the CLI with one command and no further scripting. These arguments now also have pluralized aliases: --schedulers, --sdxl-refiner-schedulers, and --s-cascade-decoder-schedulers, to help with potentially annoying typos. Using multiple schedulers contributes to generation steps combinatorically, in the same way as other diffusion parameter arguments such as --inference-steps and --guidance-scales etc.

• New Edit -> Insert Argument Console UI dialog, you can now select dgenerate arguments from a drop down and insert them into the config. This new dialog has a help button which will display the help text for the selected argument from dgenerates --help output in isolation, similar to the help window available in the image processor selection UI. This dialog is also available by right clicking the text input area in the Console UI and selecting Insert Argument.

v4.3.4 (PAG. Flux Control Nets, img2img & inpainting) Patch

v4.3.4 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.3.4 Patch

• --pag-scales, --pag-adaptive-scales, --sdxl-refiner-pag-scales, and --sdxl-refiner-pag-adaptive-scales are now properly reproduced in configuration produced by --output-configs. Previously they were left un-pluralized (typo)

4.3.0 Features

• Flux img2img, inpainting, and Control Net support. InstantX/FLUX.1-dev-Controlnet-Union is also supported with --model-type torch-flux via a mode URI argument to --control-nets, see: Flux Union Control Net Mode in the README.

• PAG perturbed attention guidance support for Stable Diffusion, SDXL, and SD3, see: Feature Table for support caveats. New arguments have been added for working with PAG: --pag, --pag-scales, --pag-adaptive-scales, --sdxl-refiner-pag, --sdxl-refiner-pag-scales, and --sdxl-refiner-pag-adaptive-scales.

• Console UI recipe form updates for Flux control nets, added basic UI for specifying one control net model, control net scale, and control net mode for Flux ControlNet Union.

v4.3.3 (PAG. Flux Control Nets, img2img & inpainting) Patch

v4.3.3 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.3.3 Patch

• --help text information fix under --text-encoders. Model type keyword T5TextModel corrected to T5EncoderModel.

• Fix for user configurable environmental variable DGENERATE_WEB_CACHE_EXPIRY_DELTA, setting value to forever no longer
  triggers a datetime overflow. DGENERATE_WEB_CACHE_EXPIRY_DELTA=forever should retain files downloaded to dgenerates
  cache forever, this cache is unrelated to the cache in HF_HOME, and is used for models / files downloaded from locations other than huggingface.

4.3.0 Features

• Flux img2img, inpainting, and Control Net support. InstantX/FLUX.1-dev-Controlnet-Union is also supported with --model-type torch-flux via a mode URI argument to --control-nets, see: Flux Union Control Net Mode in the README.

• PAG perturbed attention guidance support for Stable Diffusion, SDXL, and SD3, see: Feature Table for support caveats. New arguments have been added for working with PAG: --pag, --pag-scales, --pag-adaptive-scales, --sdxl-refiner-pag, --sdxl-refiner-pag-scales, and --sdxl-refiner-pag-adaptive-scales.

• Console UI recipe form updates for Flux control nets, added basic UI for specifying one control net model, control net scale, and control net mode for Flux ControlNet Union.

v4.3.2 (PAG. Flux Control Nets, img2img & inpainting) Patch

v4.3.2 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.3.2 Patch

Fix for --model-cpu-offload with 8 bit bitsandbytes quantized weights.

Fix for diffusers / accelerate randn_tensor device mismatch error, occasionally occurring in the following scripted construct upon subsequent invocations:

#! /usr/bin/env dgenerate --file
#! dgenerate 4.3.2

# Iterate over any scheduler options

\setp schedulers [ \
    'EulerAncestralDiscreteScheduler', \
    'KDPM2AncestralDiscreteScheduler', \
    'EulerDiscreteScheduler', \
    'KDPM2DiscreteScheduler', \
    'HeunDiscreteScheduler']

{% for scheduler in schedulers %}
    stabilityai/stable-diffusion-xl-base-1.0
    --model-type torch-sdxl
    --dtype float16
    --variant fp16
    --scheduler {{ scheduler }}
    --model-cpu-offload
    --prompts "add your prompt here"
{% endfor %} !END

Issue only occurred with --model-cpu-offload/--model-sequential-offload enabled due to quirks with accelerate offloading and high memory usage.

4.3.0 Features

• Flux img2img, inpainting, and Control Net support. InstantX/FLUX.1-dev-Controlnet-Union is also supported with --model-type torch-flux via a mode URI argument to --control-nets, see: Flux Union Control Net Mode in the README.

• PAG perturbed attention guidance support for Stable Diffusion, SDXL, and SD3, see: Feature Table for support caveats. New arguments have been added for working with PAG: --pag, --pag-scales, --pag-adaptive-scales, --sdxl-refiner-pag, --sdxl-refiner-pag-scales, and --sdxl-refiner-pag-adaptive-scales.

• Console UI recipe form updates for Flux control nets, added basic UI for specifying one control net model, control net scale, and control net mode for Flux ControlNet Union.

v4.3.1 (PAG. Flux Control Nets, img2img & inpainting)

v4.3.1 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.3.1 Patch

Include transient dependency bitsandbytes on Windows and Linux for loading quantized weights.

Documentation and --help output fixes.

4.3.0 Features

• Flux img2img, inpainting, and Control Net support. InstantX/FLUX.1-dev-Controlnet-Union is also supported with --model-type torch-flux via a mode URI argument to --control-nets, see: Flux Union Control Net Mode in the README.

• PAG perturbed attention guidance support for Stable Diffusion, SDXL, and SD3, see: Feature Table for support caveats. New arguments have been added for working with PAG: --pag, --pag-scales, --pag-adaptive-scales, --sdxl-refiner-pag, --sdxl-refiner-pag-scales, and --sdxl-refiner-pag-adaptive-scales.

• Console UI recipe form updates for Flux control nets, added basic UI for specifying one control net model, control net scale, and control net mode for Flux ControlNet Union.

v4.3.0 (PAG. Flux Control Nets, img2img & inpainting)

v4.3.0 release with Windows installer.

Due to the size of the packaged python environment, the installer is within a multi-part zip file.

The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/

Download both dgenerate_installer.zip.001 and dgenerate_installer.zip.002 to a folder.

Unzip dgenerate_installer.zip.001 to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi to install.

dgenerate will be installed under C:\Program Files\dgenerate by default with an isolated python environment provided.

The install directory will be added to PATH, and dgenerate will be available from the command line.

Portable Install

A portable install is provided via dgenerate_portable.zip.001 and dgenerate_portable.zip.002, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.

4.3.0 Features

• Flux img2img, inpainting, and Control Net support. InstantX/FLUX.1-dev-Controlnet-Union is also supported with --model-type torch-flux via a mode URI argument to --control-nets, see: Flux Union Control Net Mode in the README.

• PAG perturbed attention guidance support for Stable Diffusion, SDXL, and SD3, see: Feature Table for support caveats. New arguments have been added for working with PAG: --pag, --pag-scales, --pag-adaptive-scales, --sdxl-refiner-pag, --sdxl-refiner-pag-scales, and --sdxl-refiner-pag-adaptive-scales.

• Console UI recipe form updates for Flux control nets, added basic UI for specifying one control net model, control net scale, and control net mode for Flux ControlNet Union.