CommerceGen-Studio: An End-to-End Framework for Commercial-Grade Main-Image Generation
CommerceGen-Studio generates e-commerce main images at millions per day and leads open baselines.
CommerceGen-Studio is an end-to-end pipeline for commercial e-commerce main images: PERL-Agent distills SKU metadata into selling points, Layout Thinker Optimization aligns rewriting with rendered quality, and SKU-DiT generates from the product image, trademark crop, and layout. The authors say it runs in production at millions of images per day with latency of a few seconds. On CommerceGen-Bench, hard cases from live traffic, it scores highest among evaluated open-source baselines on both editing tracks and matches larger closed models on instruction following and product integration.
- Three stages: extraction, rewriting, and multi-condition generation.
- SKU-DiT is trained with Reinforce-Distill-Reinforce.
- Reported production scale is millions of images per day.
- Leads open baselines on CommerceGen-Bench editing tracks.
Full article221 words · extracted from huggingface.co · click to collapse
We present CommerceGen-Studio, an end-to-end framework for generating commercial-grade main images for e-commerce Stock Keeping Units (SKUs). Our framework is organized along a three-stage pipeline of information extraction, semantic rewriting, and multi-condition generation: Prompt-Eval-Refine-Loop Agent (PERL-Agent) distills heterogeneous SKU metadata into concise selling-point text lists through an iterative closed loop of VLM and LLM roles scored by a reference-free rubric. Layout Thinker Optimization (LTO) aligns the rewriter with both caption and rendered-image quality via alternating domain-specific policy optimization. SKU-DiT, trained under Reinforce-Distill-Reinforce (RDR) progressive training, conditions jointly on the white-background product image, the trademark crop, and the structured layout to hold subject fidelity, trademark integrity, and character-level text accuracy together. The framework runs in production at a scale of millions of main images per day with a latency of a few seconds per image, which we believe makes it among the first systems of this kind to operate at commercial scale. We also release CommerceGen-Bench, a benchmark of hard cases collected from live traffic across all product categories. On it, our framework achieves the highest overall scores among the evaluated open-source baselines on both editing tracks, with and without prompt enhancement. It also matches much larger closed-source models, and beats several of them at Instruction Following and Product Integration - the two properties that determine whether an image can be published.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2610.20603