- support for video editing via Bernini 1.3B and Bernini 14B with just a prompt - a new LoRA automatically added by default for Krea2 text to image that drastically improves prompt adherence - image resolution now supports 1080p - macOS support added for Krea2 text to image (was Windows only before) - image editing with a mask only now uses Krea2 for the inpaint operation; image editing with a mask and reference image still uses Flux2/Qwen Edit - support for tap powered masking (i.e. enable mask mode, tap on the spot in the image you want masked, and it will auto-mask for you) - fixed issue where resolution update in the UI did not apply to the actual generation itself - fixed UI issue that showed a duplicate queue submission for Krea2 (if Krea2 throws an error)

-for Windows, text to image now uses Krea2 Turbo for faster image creation and with a much wider creative touch than Flux2 with less prompt tinkering. Ultra Speed uses 4 steps, Speed uses 8 steps, and Quality uses 16 steps

-restored use of flux2 dev pi (4 step inference model) for ‘Quality’ text to image generation -'Quality' edit now uses a special Qwen Edit 4 step model for superior image editing, especially with scene/pose changes and keeping characters consistent between shots. Supports masking and in-paint references too! -server now has a version endpoint for FluxMotion, and the app will alert the user if their server is not the latest version (for bugfixes or missing features)

-server now correctly returns the seed used for generation (enabling you to re-use the same seed, but change other options, such as doing export quality for video) -updated auto-masking endpoint to work on mac (previously was hard-coded for CUDA) -fixed an issue where url based images from your browser were not showing the FluxMotion app when sharing -ability to re-use a seed from a previous generation, allowing you to: create a draft version of a video that you like, then keep the seed and run the ‘Quality’ preset to get a 1:1 likeness but at a higher quality -now when generating a video with another video in the preview, it will use the image for that video instead of throwing an error

-tweaked some video settings; ultra speed is 16fps 2 step, speed 16fps 4 step, and quality is 4 steps with 24fps (final export quality) -video defaults to 480p (better default quality) -app (and server) both support end frames for videos, giving you great control over the start/end of the video you want to make -quality image mode now uses the non-distilled Flux2 Klein 9b base model at 15 steps, which is similar in quality to the previous 4 step Flux2 DEV pi model used before -a new consistency LoRA was added for ‘Speed’ and ‘Quality’ which helps maintain character posing, proportions or size in an image when changing scenes -support for importing a photo/image from any other app directly into FluxMotion (bypasses the need to save to photos or files app as an intermediary step) -hiding options on partial preview results (like download/upscale, or video controls) -prevent masking/upscaling while other generation calls are active (this causes the server to crash and be overloaded)

Major feature additions: -support for automatic masking of images based on a prompt -support for text only in-painting, as well as reference image based in-painting -server now downloads models when you use them, and you see this status in the app -support for choosing a resolution to use: 480p, 720p, and a bonus 256p for video generation -video "Ultra Speed" mode -support for importing more than 1 image into the gallery at once Bugfixes: -fixed bug with how delete was handled for imported images -fixed some UI responsiveness issues -removed accidental text file saving during Files export

- you can now run the server on your mac! All features are supported except for video generation (work in progress) - demo mode now only activates if you enable it (and it is clearer that demo mode is activated)

-brand new feature; queue management. You can now remove any pending jobs from the queue at any time -fixed issue where queue was processed out of order -fixed issue where deleting a base image did not correctly cleanup variant images too, causing extra load calls with no image to show for it

- fixed major issue where error showed upon app launch

-updated to use wan2.2 lightning for image to video; produces much higher quality outputs -users no longer have to clear their prompt to see past prompts when scrolling through (less fighting with the prompt) -support for 9:16, 16:9, and 4:3 aspect ratios for images and videos