github invoke-ai/InvokeAI v7.0.0-alpha.2
InvokeAI 7.0.0 (alpha 2)
on GitHub

pre-release3 hours ago

Caution

⚠️ Please read this before you install

If you point v7 at your existing InvokeAI root directory, it will upgrade your database, and that database will no longer work with InvokeAI v6. The migrations only go one way, and v6 refuses to open a database that contains migrations it doesn't recognise.

To try v7, create a brand-new InvokeAI root directory and experiment there. In the Invoke Launcher , choose a new, empty install location. When prompted for the version, select Manual and enter version v7.0.0-alpha.2. For a manual install, pass --root /path/to/new/folder or set INVOKEAI_ROOT. Keep your v6 install and its root directory exactly where they are.

Upgraded your v6 database by accident? Don't panic. Invoke makes a backup before it migrates anything. See Restoring your v6 database at the bottom of these notes.

Warning

This is alpha software. Things will break, some features are half-finished, and a few corners still have cobwebs. Please don't trust it with work you can't afford to lose. When you hit a bug, and you will, please open an issue on GitHub and put v7 in the title, or tell us about it on our Discord. Screenshots, logs and the output of the new Diagnostics panel all help a lot.

InvokeAI 7.0.0 (alpha 2)

Since alpha 1 we've added a few features and a large number of bug fixes and stability improvements.

Major Features

Major new features include:

Multi-board projects

In the previous alpha, every Project had just one board associated with it. We have changed this so that you can now create unlimited numbers of project-specific boards. (@joshistoast )

Support for high performance database backends

You can now swap out the default sqlite3 database backend for high-performance databases MySQL and MariaDB (@Pfannkuchensack ). This is particularly useful for multi-user installations. See MySQL and MariaDB for information on configuring this option.

Support for more GGUF format transformers.

We now provide support for krea2 GGUF Q8_CR transformers models, ideogram4 GGUF transformers, and ernie-image GGUF transformers. (@Pfannkuchensack )

Improved video support

The external remote control API (used, for example by PhotoMapAI) now allows an external program to send reference videos and first/last frame images to the Video panel. In addition, the MiniMax H3 Ref2VA model will now run when given text prompts and audio-only references (@lstein)

Atlas Cloud image provider

The Atlas Cloud image provider has been added to the list of externally hosted (API) models that you can call into from InvokeAI (@binyangzhu000-sudo).

New boolean workflow node

Thsi version adds a simple boolean logic node to performs boolean AND, OR, XOR, NOT, NAND, NOR, or XNOR operations on inputs A and B. We also fixed a bug with the If node, that could cause deadlocks in some circumstances (@JPPhoto )

New configuration option for VRAM

To address OOM errors, we now provide a new invokeai.yaml tuning option, reserve_vram_gb, that will subtract a configured amount from the model cache's available VRAM budget. This will reduce the VRAM reserved for model loading by the indicated amount and let it be used for intermediate storage during generation. The default is 0.0, preserving existing behavior unless configured. (@JPPhoto )

Major Bug Fixes

Notable bug fixes and performance improvements include:

  • Reduce the size of the imagemap index by storing its values as 16-bit integers (@JPPhoto )
  • Let Expand Prompt and Image to Prompt be reopened while they run (@unmarco)
  • Multiple stability fixes to the model manager to address partial and corrupted downloads (@JPPhoto )
  • Make PiD 1.5 usable for FLUX.1 and FLUX.2 (@Pfannkuchensack)
  • Update all dependencies that are known to have high or critical security issues (@lstein )
  • Version 10 of the ROCm libraries for improved performance on AMD GPUs (@Pfannkuchensack ).
  • Reference audio files that contain multiple tracks are now handled properly. Previously you could end up with distorted audio if the tracks had different sampling rates. (@agammann )

The full list of updates follows.

What's Changed

  • chore(release): bump version post 7.0.0-alpha.1 by @lstein in #9675
  • [SD] Modular backend - add ip adapter by @StAlKeR7779 in #9678
  • Fix arguments for hidiffusion hook by @StAlKeR7779 in #9679
  • feat(external): add Atlas Cloud image provider by @binyangzhu000-sudo in #9476
  • Preserve nested Iterate value scope by @JPPhoto in #9677
  • Admin database compaction by @JPPhoto in #9688
  • Update README with new backers and badge format by @lstein in #9701
  • docs: add policy for abandoned community PRs by @lstein in #9687
  • docs: add v7 known bugs for multi-user sharing and the delete-confirmation opt-out by @lstein in #9672
  • chore(CI/CD): relax codeowners by @lstein in #9647
  • Store image-index embeddings as fp16 by @JPPhoto in #9696
  • fix(webv2): let Expand Prompt and Image to Prompt be reopened while they run by @unmarco in #9689
  • feat(krea2): load ComfyUI-GGUF Q8_CR transformers by @Pfannkuchensack in #9697
  • Revalidate collector dependents on input changes by @JPPhoto in #9686
  • feat(video): let LTX-2 soundtrack conditioning keep first and last frames by @lstein in #9648
  • Add gallery maintenance UI by @JPPhoto in #9700
  • feat(webv2): add configurable donation links by @joshistoast in #9651
  • feat(ui): reliability pass plus canvas, gallery and frontend layout features by @joshistoast in #9682
  • ci(docs): deploy main's docs to v7.invoke.ai by @lstein in #9704
  • feat(video): place a gallery image in the Video panel from the recall API by @lstein in #9693
  • feat(minimax-h3): allow audio-only Ref2VA reference sets by @lstein in #9684
  • Fix model manager cache and recovery failures by @JPPhoto in #9669
  • fix(external models): accept declared ratios in any terms and keep snapped sizes in bounds by @lstein in #9681
  • fix(model-cache): reserve the GGUF dequantization transient in every node by @Pfannkuchensack in #9709
  • feat(ideogram4): load community GGUF transformers by @Pfannkuchensack in #9708
  • Gallery: Browse large collections with a sparse page cache (infinite scrolling) by @JPPhoto in #9683
  • fix(deps): resolve high and critical dependabot alerts by @lstein in #9723
  • chore(deps): bump diffusers to 0.41.0 by @Pfannkuchensack in #9774
  • feat(db): move every query into one database layer and support MySQL and MariaDB by @Pfannkuchensack in #9702
  • Add reserved VRAM budget option by @JPPhoto in #9743
  • feat(ltx2): auto duration for video extensions by @lstein in #9671
  • fix(pid): make PiD 1.5 usable for FLUX.1 and FLUX.2 by @Pfannkuchensack in #9772
  • feat(deps): ROCm 10 on Linux and Windows from AMD's index by @Pfannkuchensack in #9643
  • test(workflows): fix flaky last_run_at listing test on Windows by @lstein in #9788
  • fix(model-manager): prevent stale in-flight model loads from reentering the cache by @lstein in #9694
  • Add Boolean logic invocation by @JPPhoto in #9781
  • docs(models): add a format support matrix for every model type by @Pfannkuchensack in #9726
  • Fix scheduler deadlock with nested If nodes using literal conditions by @JPPhoto in #9786
  • feat(ernie-image): load GGUF transformers by @Pfannkuchensack in #9725
  • fix(audio): select first stream for PCM extraction by @agammann in #9721
  • fix(deps): resolve high and critical npm dependabot alerts by @lstein in #9783
  • feat(boards): boards belong to projects, with a Library tier and whole-project transfer by @joshistoast in #9713

New Contributors

Full Changelog: v7.0.0-alpha.1...v7.0.0-alpha.2

Check out latest releases or
releases around invoke-ai/InvokeAI v7.0.0-alpha.2

Don't miss a new InvokeAI release

NewReleases is sending notifications on new releases.

Get notifications