A first concept
A designer can turn a rough thought into a brief that is easier to discuss. Use it to identify the framing, atmosphere and words you want before spending time generating versions.
Shape a short idea into a structured creative direction.
A thought, made specific. Develop a short idea into a creative brief without uploading an image.
0/2,000 characters. Quote any exact wording you want in the image.
Up to 5 connected attempts per tool each UTC day. Remaining usage appears after a request.
Processed in memory. Uploads are never saved by this site. Privacy details
Your prompt will appear here, ready to edit, copy, and make your own.
Describe the subject and what you want to make: a poster, photograph, illustration or another image. Start with the detail that matters most.
Pick a style and aspect ratio. Keep GPT Image 2 as the default writing target or try GPT Image 2.5. Quote exact wording if the image should contain text.
Generate, read the added details and revise anything that does not fit. Copy the edited brief into a compatible image tool to create the image separately.
This original writing example is not a live analysis of an uploaded reference.
An original idea: a quiet neighborhood café poster with the words ‘SUNDAY SLOW’ at the top. Example artwork pending: TODO:用生图工作台生成。
Create a restrained café poster with a softly lit window scene as the main visual. Leave generous space above the subject. Place the exact words ‘SUNDAY SLOW’ at the top in a clear, high-contrast headline. Use a quiet palette and natural surface texture. Keep the layout simple and add no other text.
Organize the idea into subject, scene, composition, light, materials, colors and text. The fields make decisions visible without claiming a special image-model syntax.
Choose whether lettering belongs in the image. Quote the words you want, then review their placement and check the rendered spelling when you generate elsewhere.
Added details are suggestions. Edit the result or ask for a different phrasing, keeping the parts that support your idea and removing unnecessary instructions.
A designer can turn a rough thought into a brief that is easier to discuss. Use it to identify the framing, atmosphere and words you want before spending time generating versions.
A maker or collaborator can read the proposed choices and replace the ones that do not fit the project. The brief supports a discussion; it does not establish product facts, usage rights or guaranteed visual results.
An idea becomes easier to develop when it names the kind of image you need. A poster needs room for a message, while a product photograph may place more emphasis on shape, surface and light. Tell the generator what you are trying to make instead of relying only on a mood word. The tool can suggest supporting details, but you remain responsible for deciding whether they serve the project. A short, concrete starting point is often easier to revise than a long list of unrelated visual wishes.
Identify the main subject and the relationship you want to show. A cat sitting beside a café window is a clearer starting point than several disconnected style labels. You can then decide how much of the setting belongs in the frame and where the viewer’s attention should go. The generator expands the idea into a proposed arrangement; it has no reference image to verify those choices against. Treat additions as creative suggestions rather than observations, and remove any detail that distracts from the subject you actually meant.
The style selector offers a small set of visual directions. Use one as a starting point, then check whether the written description supports it with appropriate materials, color and lighting. Avoid stacking incompatible styles just to make the brief sound elaborate. If you are unsure, keep the style broad and focus on a readable composition. The target selector separately chooses GPT Image 2 or GPT Image 2.5 as a writing preset. It does not choose the backend vision model, a provider account or an image-generation endpoint.
If lettering is part of the design, enable text and quote the exact phrase in your idea. Decide whether it is a headline, a label or a small supporting line, and leave enough room in the composition for it to be readable. If text is disabled, the brief asks for no added lettering. Inspect the written result before copying and check the actual generated image afterward. Neither the checkbox nor a precise quotation guarantees perfect rendering. Do not leave important wording to be inferred from a vague description of the topic.
A tall poster, a square product tile and a wide scene invite different arrangements. Choose the shape that suits the place where you expect to use the image. The aspect-ratio preference appears as a creative instruction in the brief; it is not a claim that every receiving image tool exposes the same size control. Confirm dimensions separately in that application. If you change the ratio later, review subject placement, cropping and text space together rather than assuming that the previous composition will adapt without another decision.
Read the finished brief once for intent and once for contradictions. Is the subject still the one you meant? Does the lighting fit the setting? Is the requested text exact? Simplify anything that competes with the main idea, and keep a copy of the version you like in your own workspace. Aizvu returns editable text or JSON; it does not generate the image or keep a project history. Local example mode demonstrates the interface without analyzing or expanding your idea through a live provider, and its sample label should remain part of how you interpret the result.
No. This tool starts from a written idea. Use the image-to-prompt tool when you want a visual reference analyzed instead.
GPT Image 2 is the default and GPT Image 2.5 is optional. These choices change the writing preset; the configured provider remains separate.
No. They are creative suggestions based on your idea. Review them and replace any assumption that does not belong in your project.
Enable text, quote the exact words in your idea, and explain their role or placement. If you disable text, the brief asks for no added wording.
Yes. Detailed, short and JSON formats are available. JSON fields are Aizvu’s editorial structure rather than a required vendor schema.
It returns prompts and does not generate images. There is no uploaded reference in this mode; connected requests send the idea to the configured provider, whose policies apply.