Image models just learned the two words designers have been begging for
image models just learned the two words designers have been begging for: leave it. black forest labs shipped flux 3 image on oct 1 with bounding-box placement, up to ten reference images and native 4k, doing multi-turn edits that keep untouched pixels intact.
the regenerate-and-pray era is ending.
Context
Black Forest Labs introduced FLUX 3 Image on October 1. Its docs describe generation from a prompt, edits by written instruction, up to ten reference images, and bounding boxes to place each element, with resolutions up to 4k.
Bounding boxes can also limit an edit to one element so the rest of the image stays unchanged, per the docs. The Decoder reported the multi-step editing that leaves the rest of the picture alone.
'Keep untouched pixels intact' is the vendor's claim about edits. The docs say the rest stays unchanged when a box targets one element. No independent pixel-level test was found.
References are 1 to 10 images, each 256 x 256 px up to 16 MP. 'Native 4k' matches the resolution setting.
'The regenerate-and-pray era is ending' is the author's opinion.
Related work
- Black Forest Labs: FLUX 3 Image docs ↗Endpoint, references and resolution options.
- Bounding boxes with FLUX 3 Image ↗How boxes place and edit single elements.
- The Decoder on the launch ↗Independent report, Oct 2.
Watch next
- Look for third-party tests of how well untouched regions survive repeated edits.
Sources
- BFL docs: FLUX 3 Imagedocs.bfl.ml
- BFL docs: bounding boxesdocs.bfl.ml
- The Decoder, Oct 2, 2026the-decoder.com
- Launch Archive: FLUX 3 Image, Oct 1, 2026launcharchive.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 9 October 2026 at 18:39 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →