Image models are good at light and set dressing and bad at type, Arabic above all. Adshot splits the job. The model builds the scene around the product and leaves room for the words; the headline, the call to action and the logos go on afterwards as layers on a canvas, so the type stays sharp and the Arabic shapes correctly.
- Each upload gets a role, product, style, composition, background, logo or prop, which tells the model how to treat it.
- A brief sets purpose, mood, ratio, market and language, with art direction such as a water splash, marble or neon.
- A vision model reads the product before the render, so the prompt describes what was actually uploaded.
- Backgrounds are cut out in the browser by an on-device model.
- Finished posters export at 9:16, 1:1, 16:9 or 4:5.
- Arabic and English as real layers
- Four ratios
- 200 AI credits a month