English summaries of our Japan-market benchmark series. Full outputs, prompts and cost details are available in the original Japanese articles.
A cross-modal benchmark series using the traditional Japanese uchiwa fan as a consistent subject across T2I, I2I and I2V.
NanoBanana 2, FLUX.2 [dev] and GPT Image 2 compared on Japanese uchiwa (summer fan) text-to-image generation.
Exploratory I2I benchmark replacing a product-photo uchiwa background with a Japanese tatami room.
Veo 3.1, Veo 3.1 Lite, Kling 3.0 V3 Standard and Seedance 2.0 compared for uchiwa I2V generation.
GPT Image 2, NanoBanana 2 and FLUX.2 [dev] compared on Japanese businesswomen in meeting, cafe and office scenes.
Veo 3.1, Veo 3.1 Lite, Kling 3.0 V3 Standard and Seedance 2.0 compared for property and landscape video.
GPT Image 2, NanoBanana 2 and FLUX.2 [dev] compared across six age/gender groups in a Tokyo office.