Changelog.

What changed, release by release.

1.3.8Nov 5, 2026
  • A full macOS menu bar: Genvas › Check for Updates…, Settings… (⌘,), File › New Board (⌘N), Boards, Projects, Assets and Stats on ⌘1–4, and Help links to the site, support, terms and privacy.
  • Check for Updates opens Settings › About and checks straight away.
1.3.7Nov 5, 2026
  • GPT Image 2.6 Flare and Sunburst on OpenAI: a Complexity setting, plus SVG and animated GIF output.
  • Agent runs with two or more GPT Image 2.5 jobs go through OpenAI Batch at half price; results usually arrive in minutes, at most 24 hours.
  • Draft runs only lower the nodes where a draft is cheaper; everything else runs as set, and the price check shows which.
  • New video models: FLUX 3 Max keyframes, MiniMax H3 Max, H3 Max Turbo and H3 Max Relight on fal, and Seedance 2.5 Draft with a 1080p upgrade of the draft you keep.
  • Meshy multi-rig packs keep every animation clip in the mesh’s history, and the empty board’s Add a node card fits 3D.
1.3.6Nov 3, 2026
  • New full-screen viewers for images, audio, video and 3D: open any card to crop, split sheets, upscale, trim audio and video, and combine clips with audio or images, with live prices on paid tools.
  • Image and video viewers zoom with pinch or ⌘-scroll at the cursor, pan by dragging, and toggle Fit and 100% with a double-click or the + − 0 1 keys.
  • 3D viewers for models, animations, CAD, PLY and STL show what each model was made from, zoom toward the cursor and orbit any point you double-click.
  • Cancelling a paid tool now cancels the provider job where the provider allows it, and says whether anything was billed.
  • Board 3D and media cards show clip length, size and file type up front, and missing files can be found again or shown in Finder.
1.3.5Nov 2, 2026
  • Expanded image, video, audio, 3D and text models across fal, kie, Sume, ElevenLabs v4, Fish Audio and Gemini, with dead-key checks in Settings.
  • Local CAD and 3D conversion (STEP, IGES, BREP, GLB, STL, OBJ), multi-format 3D exports, mesh segmentation, multi-animation rigging, PBR material maps, layered RGBA decomposition and depth maps.
  • Speech transcription with SRT captions, video-to-music scoring, text-to-SVG vectorization, 6-stem audio separation and local video composition with voiceover, background music and burned-in captions.
  • Group layout cleanup, portrait stack spacing, sound card downloads, animated FBX previews and canvas zoom performance improvements.
1.3.4Oct 7, 2026
  • Agent guidance favors one sheet for matching asset sets and native transparency where supported.
  • Board warnings flag separate asset jobs and missing transparency.
  • Agent additions automatically clean touched groups while preserving explicit node positions.
1.3.3Oct 5, 2026
  • 3D models: generate a 3D model from a prompt or reference images (fal), shown on a new 3D card with a live GLB viewer you can rotate.
  • 3D tools work on an existing model: rig and animate, retexture, remesh, smart topology, SAM 3D, TripoSplat and Hunyuan Motion.
  • 3D cards, gallery and history show automatic preview thumbnails, and downloads keep the right file type (GLB, FBX, PLY).
  • The bundled agent tools can build and run 3D stacks, and check 3D inputs before changing the board.
  • A node that can't run yet (a missing wire or input) now says so, and that nothing was sent to the provider, instead of blaming the provider.
1.3.2Oct 2, 2026
  • Split sheet can cut sheets into an even grid of columns and rows. Clean up packs spread-out groups and pulls distant refs back into the group.
  • Connect in the app sets up the bundled MCP tools and prompting skill for Claude Code or Codex; the DMG no longer needs a separate setup step.
  • Agent model discovery shows models supported by configured provider keys, with unavailable models available on request. Direct generation checks for a supporting key before submission.
  • Price quotes and spend history better account for request settings and reported usage. Unpriced or incomplete costs stay marked as unknown or estimated.
  • You can delete the last board and create a new one afterward.
  • A downloaded update stays ready to restart, and an updated notice links to What’s new after the app restarts.
  • The spend summary is less crowded, and video playback controls sit above a continuous fade.
1.3.1Oct 2, 2026
  • Groups have a Draft button that runs every draft-capable image node at 1k / low, so you can check a group cheaply before the full run.
  • About is now a title card; these notes moved to their own What’s new view.
  • Board tabs give way before folders, a folder's boards open together, and the More menu stays inside the window.
  • Agent activity shows provider logos, run results and out-of-balance notices. Agents use the provider you picked in the app, and OpenAI's own image models default to OpenAI.
  • The bundled agent tools add workflow guides for ads, film, UGC, game assets, audio and continuity, plus free local editing tools. Layout guides can be added from + and right-click.
  • Selecting or panning no longer saves the board. Fixed run notices, provider error messages and model packaging.
1.3.0Oct 1, 2026
  • Agent activity now shows brief start, done, and error notices with links to the relevant board or node. Notices stay quiet on reconnect.
  • The bundled MCP tools and prompting guide now agree on model options and guide creation. Guide cells survive generation and editing.
  • Fixed comment switching that could overwrite another comment, plus collaboration and media handling issues found in the audit.
  • Restarting for an update now waits for board changes to save and reports a failed save; a failed install keeps the downloaded update ready to retry.
  • Improved request logging and isolated MCP test data from desktop sessions.