{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/a37bf74aa4934085a56ecc572b49ea9f\" frameborder=\"0\" width=\"1726\" height=\"1294\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1294,"width":1726,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1294,"thumbnail_width":1726,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/a37bf74aa4934085a56ecc572b49ea9f-5af4b640062128dd.gif","duration":120.490667,"title":"Can AI Write Urdu ? ","description":"This Loom tests whether AI image models can generate Urdu design work that matches what Kashmiri businesses actually pay for. Rukhaya used six real job prompts and generated 18 images once using three models, then had native readers evaluate them via a Google form shared on WhatsApp, considering text readability, legibility, cultural fit, and preference. Urdu was framed as India’s most unserved script in AI, and prior benchmarks rank it worst for speech models. The results showed Gemini 3.1 outperforming the rest, with Gemini 2.5 being realistic but not readable to native readers."}