One image in seconds, from a plain description. Posters, product shots, characters, backgrounds. Every engine on one balance, and a character you can save and bring back to the next image.
Sign in with a Google account and open the image studio. Pick an engine: for an image with Hebrew wording, GPT Image 2 or Grok Imagine Image 2.0, and for an image without text, whichever engine suits the style. Write what you want to see, the way you would brief a photographer: who or what is in the frame, where, in what light and in what style. The price is shown before the click, and a few seconds later the image is ready to download. Not happy with one detail? Describe the change in plain words and get a corrected version of the same image, without starting over.
An image model does not write text, it draws it as a shape it has seen before. Most of them saw billions of images with English text and very few with Hebrew, and they learned a left-to-right writing direction, so Hebrew comes out reversed or as gibberish. That is not a glitch and no phrasing fixes it. What has changed: two engines in the system, GPT Image 2 and Grok Imagine Image 2.0, do write Hebrew correctly inside the frame, and those are the ones to reach for when the wording has to be part of the picture. In every other case the approach that works is to generate a clean image with no words and add the caption on a separate layer with a real font. That comes out sharper anyway, in the font you picked, and editable tomorrow.
This is the difference between an experiment and content you can sell. Generate a character once, save it, and bring it back into every new image. That is how you build a post series, a campaign, or an illustrated story where the hero looks identical in every frame. Without it every image produces a different person, which is exactly what makes content look like AI.
A good description answers four questions: what we see, where, in what light, and in what style. For example: "a cappuccino on a wooden table in a Tel Aviv cafe, soft morning light from the window, steam rising from the cup, close-up in the style of product photography". Every detail you leave out, the engine invents on its own. And if the image came out almost right, change one detail and generate again instead of rewriting everything.
A product photo on a new background with no photographer or studio, a professional portrait for a profile and a website from a single selfie, social posts with a Hebrew headline inside the image, and a banner for a sale. Upload an existing photo of the product or of yourself, describe the image you want, and get it within seconds. Every image is saved in your gallery and can go straight into a video.
Every image costs computing power, so there is no truly free and unlimited. What there is: an image starts at one credit, and the price of every image is shown before you click. You buy a credit pack, the same balance works on every engine, and an image that fails goes back to your balance automatically.
An image starts at one credit; the price depends on the engine and resolution and is shown before you click. An image that fails goes back to your balance.
Yes. Write what you want to see, the way you would brief a photographer, and the engine understands the intent.
Because most models draw letters as shapes and learned a left-to-right direction from hundreds of millions of examples, and no prompt forces them otherwise. Two engines in the system do solve it, GPT Image 2 and Grok Imagine Image 2.0. For all the rest, generate the image without text and add the caption on a separate layer.
Yes. Upload the image and describe the change in plain Hebrew, with no tools and no layers.
Home · Pricing · What's new · All engines · All effects · Help center · Lessons · Meital Sela · AI video · AI images · AI voice · AI music · Image to video · What it costs · Credits, not a subscription · Blog · Connect to Claude and ChatGPT · Presets · Privacy · Terms
Cadabra AI · Meital Sela · Rehovot, Israel