LoginSign Up
← Back to all sparks
I

InvokeAI

AI-ASSISTANTS
Velocity6.3

Open-source generative AI image creation and workflow studio

invoke.com ↗

InvokeAI 7 reaches beta with projects and a single widget workbench as the new default.

image generationopen sourcev7 rewriteworkbench uimodel support
◆Current state
InvokeAI 7.0.0 beta 1 shipped on November 9 after four alphas, making the rebuilt interface the default. v7 organizes work into autosaving projects that hold canvas, settings, workflows and boards, and replaces separate tabs with one workbench of movable widgets, including Video. It requires Python 3.12 and a one-way database migration, so v6 users are told to install into a fresh root.
◆Where it's heading
The v7 rewrite has moved from tester alphas to a feature-complete beta, and the v6 line is down to small patches. The next stretch is bug fixing and moving v6 users across, with the legacy interface kept only as a fallback.
◆Prediction
Expect further v7 betas, then a 7.0.0 release, with v6 limited to critical fixes like the 6.14.4 startup patch.

◆Recent moves

  1. 4d ago

    InvokeAI 7 beta makes projects and the new workbench default

    ⚡ SPARK

    Beta 1 closes the alpha phase of the v7 rewrite and makes the new interface the default, which turns the two-line split into a migration. It needs Python 3.12 and a database upgrade v6 cannot read.

    View source ↗
  2. 5d ago

    InvokeAI re-records its page-size performance baselines

    An internal CI change that accepts slightly larger launchpad and editor bundles. It is the performance budget the v7 notes describe, with no user-facing change.

    View source ↗
  3. 11d ago

    InvokeAI 6.14.4 fixes a startup crash after upgrading

    A one-fix patch that caps the Starlette library so the web UI loads again for users who upgraded to 6.14.3. It is cleanup on the v6 maintenance line while new work goes to the v7 alphas.

    View source ↗
  4. 12d ago

    InvokeAI v7 alpha 4 adds FLUX.3 Klein and MOSS-Image

    Alpha 4 adds FLUX.3 Klein and MOSS-Image, edit references for Krea-2 and Qwen-Image-2.1, better live previews and Windows on ARM64. It is the largest v7 alpha so far, but still pre-release and gated behind a one-way database migration.

    View source ↗
  5. 12d ago

    InvokeAI 6.14.3 adds editable vector layers to Canvas

    A v6 patch that adds editable vector layers to the Canvas, which can be rasterized for img2img, and fixes restarting interrupted model downloads. It shows v6 still getting user-facing features while v7 is in alpha.

    View source ↗
  6. 26d ago

    InvokeAI v7 alpha 3 adds role-based board sharing

    Alpha 3 brings role-based board sharing, a new image model, a faster NVIDIA attention backend and more restored legacy features. Board roles are the clearest sign that v7 targets teams.

    View source ↗