FLORA Weekly
No. 088, (latest issue)
Skills for Every Job
This week, your agent got a whole library of FLORA skills, and FLORA got a light mode. GPT Image 2.6, VEED Dubbing and Topaz Wonder 4 join the model lineup, admins can pin projects for their whole workspace, and big projects duplicate in the background, alongside a long list of fixes.
New
Light mode
Choose Light, Dark or System in Settings → Preferences → Appearance, or from the sidebar footer.
MoreLess
The canvas, editors and menus all follow, and System switches with your computer's setting. Dark stays the default until you change it.
One-job skills for your agent
39 new FLORA skills, including background remover, studio shot, relighting, storyboard maker, mood board, product spin and sketch to campaign.
MoreLess
Ask Claude or ChatGPT for the job and it finds the skill, asks for the right input and quotes the cost before it runs.
GPT Image 2.6 Sunburst and Flare
Two new image models for text-to-image and image editing: Sunburst for detailed work, Flare when you want speed.
VEED Dubbing and Dubbing Pro
Translate the speech in a clip into another language, with a new voice and matching lip-sync.
Topaz Wonder 4
A new upscaler, now behind the toolbar's Upscale button for sharper quick upscales.
Workspace pinned projects
Admins can pin up to 12 projects to a Pinned projects section on the Projects page that every member sees.
MoreLess
Right-click any project and choose Pin to workspace or Pin to sidebar.
Improved
Duplicate big projects
Duplicating a large project now runs in the background.
MoreLess
The copy opens right away as Copying…, finishes even if you close the tab, and offers Retry if something goes wrong.
Richer SVG import
SVGs brought into the Image Editor keep more of their look as editable vector layers, including dashed strokes, fill and stroke opacity, and translucent groups.
Faster frame extraction
Extracting the first, last or a timestamped frame from a video now returns an image in a second or two.
Uploads that survive a blip
Image, video, document and 3D uploads now retry and resume after a short network drop instead of failing.
Sharper Seedance 2.5 references
Seedance 2.5 now receives your original reference images and full-quality versions of .mov and .webm clips, for better-looking results.
Skills gallery polish
Every skill now has thumbnail art, and a skill's button changes to Added once you've added it, so you can see what's active.
Admin control over API keys
Organization admins can turn on Only admins can create API keys to stop members creating new keys across the organization's workspaces.
Fixes
- Public share links for media load again instead of showing an error.
- Prompts longer than a model's limit are now refused with both lengths shown, instead of being cut off without warning.
- An image node in Auto mode now follows the aspect ratio of its connected input image instead of defaulting to square.
- Cmd+E and right-click Rename always open the rename field, even right after extracting variants, and Cmd+Shift+D now downloads instead of duplicating.
- Clicking a link in a short note shows the full Open external link? dialog instead of a cropped one.
- New images and videos on deck slides get a box that matches their shape, so clicking beside an image no longer selects it.
- Downloads from shared deck links work again for decks that use uploaded fonts.
- Deck PDF and PNG exports skip an image that can't load, with a warning, instead of failing.
- The Create Skill menu in text, image and video nodes stays open when you click it.
- PDF pages with certain gradients render in their real colors instead of hot pink in document nodes.
- The FLORA skills gallery lists skills in ChatGPT again.
- The Refresh and What's new buttons in the update notice have proper padding again.
- The transparency checkerboard in the Image Editor shows square cells again.