Adobe Firefly Image 5 brings support for layers, will let creators make custom models (techcrunch.com)

Adobe today unveiled Firefly Image 5 and a slate of platform upgrades that push its generative tools toward professional, production-ready workflows. The new Image 5 model generates native images up to 4 megapixels (previous gen produced 1MP and upscaled to 4MP), improves human rendering, and introduces layer-aware, prompt-based editing that treats objects as editable layers (resize, rotate, prompt edits) while preserving image fidelity. Adobe also expanded Firefly’s ecosystem — supporting more third‑party models (OpenAI, Google, Runway, Topaz, Flu, etc.), added video timeline/layers editing (private beta), and introduced AI audio features (speech and soundtrack generation via ElevenLabs). Technically notable is the ability for creators to build custom image models from their own assets: a closed beta lets users drag-and-drop images, illustrations, and sketches to train style-specific models, enabling consistent brand or artist styles without manual finetuning. The Firefly web app was redesigned to let users switch between image/video generation, pick models, change aspect ratios, and access recent files and app shortcuts — streamlining iterative, multimodal creative workflows. For the AI/ML community, these moves signal a focus on higher-resolution native synthesis, compositional control through layer semantics, tighter model interoperability, and democratized fine-tuning for creatives — with implications for IP, style ownership, and competitive productization across the creative tools market.