This is a Release Candidate. We strongly suggest backing up your database before testing to prevent data loss in case of any issues.
Please let us know if you run into anything unexpected.
What's New
RC1 supported the setting skip_model_hash in invokeai.yaml. In RC2, this is replaced by a more flexible setting hashing_algorithm.
New Model Manager
The model manager is rewritten in v4.0.0, both frontend and backend. This builds a foundation for future model architectures and brings some exciting new user-facing features:
- Queued model downloads
- Per-model preview images
- Per-model default settings - choose a model’s default VAE, Scheduler, CFG Scale, etc.
- User-defined trigger phrases for concepts/LoRAs and models - access by typing the
<key in any prompt box - API key support for model marketplaces
Model Hashing
When you first run v4.0.0, it will take a while to start up as it does a one-time hash of all of your model files.
Do not panic.
Hashes provide a stable identifier for a model that is the same across every platform.
If you don’t care about this, you can disable the hashing using the hashing_algorithm setting in invokeai.yaml.
Canvas Improvements
The canvas uses a new method for compositing called gradient denoising. This eliminates the need for multiple “passes”, greatly reducing generation time on the canvas. This method also provides substantially improved visual coherence between the masked regions and the rest of the image.
The compositing settings on canvas allow for control over the gradient denoising process.
Major research & experimentation for this novel denoising implementation was led by @dunkeroni, and @blessedcoolant was responsible for managing integration into the canvas UI.
Bonus: Invoke Training (Beta)
As of v4.0.0, all references to training in the core invoke script now point to the Invoke Training Repo. Invoke Training offers a simple user interface for:
- Textual Inversion Training
- LoRA Training
- Dreambooth Training
- Pivotal Tuning Training
You can learn more about Invoke Training at https://github.com/invoke-ai/invoke-training
Minor UI/UX Enhancements:
- Canvas Brush Size Scroll can now be inverted (Thanks @joshistoast!)
- Images in the Canvas Staging Area can now be discarded individually (Thanks @joshistoast!)
- Many small bug fixes and resolved papercuts
Installation and Upgrading
To install or upgrade to version 4.0, download the zip file from the release notes ("Assets" section), unpack it, and follow the installation instructions. For upgrades, select the same installation location.
💻 Developer Changes
v4.0.0 is versioned as a major release due to breaking changes:
- The internal nodes API has been refactored to provide a stable public API. Node authors should review the migration guide.
- The internal graph execution engine is drastically simplified, resulting in more efficient and performant processing. This carries on from the changes in v3.6.0 in which graphs are no longer stored in the database.
Contributing
As a community-supported project, we rely on volunteers and enthusiasts for continued innovation and polish. Everything from minor documentation fixes to major feature additions are welcome. To get started as a contributor, please refer to How to Contribute or reach out in #dev-chat on Discord!
What's Changed
- feat(ui): Add image size badge to gallery images by @rohinish404 in #5632
- Update l2i invoke and seamless to support AutoencoderTiny, remove att… by @brandonrising in #5936
- updated model manager to display when import item is cancelled by @chainchompa in #5924
- fix(ui): disable trigger phrase form if empty by @maryhipp in #5942
- ci: add
always_runinput to checks & tests, use this on release workflow by @psychedelicious in #5929 - added hf models import tab and route for getting available hf models by @chainchompa in #5891
- docs: misc updates by @psychedelicious in #5930
- chore: cleanup DepthAnything code by @blessedcoolant in #5945
- fix(ui): force CustomSelect's to rerender when models load to update invalid state by @maryhipp in #5949
- (ui): fetch TIs on socket connect by @maryhipp in #5950
- fix(ui): control adapter errors by @psychedelicious in #5953
- chore(ui): cleanup translations by @psychedelicious in #5946
- feat: ✏️ rename "Workflow Editor" tab label to "Workflows" by @joshistoast in #5955
- feat(mm): faster hashing for spinning disk HDDs by @psychedelicious in #5952
- feat(nodes): defaults dimensions for main models by @psychedelicious in #5931
- Work around missing core conversion model issue by @lstein in #5947
- Restore ability to install models from command line by @lstein in #5948
- fix(nodes): depth anything processor (#5956) by @psychedelicious in #5961
- feat(ui): single getModelConfigs query by @psychedelicious in #5962
- fix: model load events by @psychedelicious in #5959
- docs: update CONFIGURATION.md by @psychedelicious in #5963
- fix(nodes,ui): metadata recall by @psychedelicious in #5957
- Run typegen, update version to 4.0.0rc2 by @brandonrising in #5965
Full Changelog: v4.0.0rc1...v4.0.0rc2