Mobile Comfy Creator

More models, each with the settings it was made for — and edit photos by describing the change. New models on ComfyUI • Edit with instruction: pick a FLUX.1 Kontext model, choose a photo and type what to change ("make the sky purple, keep everything else"). No mask needed. • FLUX.2 (dev, klein 4B / 9B and klein Base) and Chroma1-HD now create, remix and edit. • Split-file models: the picker lists diffusion models next to your checkpoints, grouped by family, and sets up FLUX.1 (dev, Krea, schnell), Z-Image Turbo and Base, SD 3.5 and Qwen-Image for you. Text encoders and VAEs are found on your server automatically. • Wan 2.2 5B video: text to video and image to video on the Video screen, with the right sizes and defaults. • Each model shows whether it's ready, or exactly which file it still needs — and whether your server's GPU can run it. Per-model settings • Give any model its own steps, prompt strength, sampler, size, Avoid text, LoRAs, prompt tags and more. • They sync to your other devices through iCloud (turn off under Cloud Sync), and you can export and import them. • Better starting settings for FLUX, Z-Image, SD 3.5, SDXL Lightning / Turbo / Hyper and SD 1.5. Pony and Illustrious now use CLIP skip 2, as they were trained. A prompt assistant that knows your model • Enhance and Describe an image write tags for Illustrious, Pony and SD 1.5, plain sentences for FLUX, SD 3.5, Qwen-Image and Z-Image, and short instructions when you edit with Kontext. • Tips under the prompt and one-tap fixes in Expert when a setting doesn't suit the model. Fixes • A missing model file, text encoder or node gets one clear message before anything is sent. • LoRAs now apply to Remix and Edit on ComfyUI. • The Sampler and Scheduler lists match what your ComfyUI server offers. • The Video screen no longer asks for VideoHelperSuite when your server doesn't need it, and its Wan 2.2 pickers only offer text-to-video models. • Many smaller fixes.