Model summary
Qwen image-edit (plus) with LoRA support; positioned as a strong image-editing alternative, benefiting from improving community LoRAs/fine-tunes (e.g., relighting and next-scene prediction).
Qwen-Image Edit Plus is a 20B-parameter MMDiT-based image editing model focused on advanced, prompt-driven edits with strong adherence to visual references, including LoRA-based customization. It performs especially well on product placement tasks and multi-angle product views, closely matching the reference product’s appearance. In testing, it follows prompts accurately and preserves product identity and details across angles. However, it tends to generate products that are slightly oversized relative to realistic scenes, and surfaces like clothing or fabrics can appear overly smooth/airbrushed and less photo-realistic. It is best suited for high-quality conceptual visuals and marketing-style imagery rather than perfectly natural, documentary realism.
