Start with one clear idea. A quiet room, a product photograph, an imaginary landscape—your first image does not need a perfect prompt or a page of settings.
Naldar helps you develop the idea in conversation. Naldar Studio runs supported image models on your Mac. For a first attempt, concentrate on the subject, the setting and the light.
A simple first goal
Create one image with a compatible installed model. Look at the result, identify one thing to improve and change that one thing. A small, repeatable experiment is easier to learn from than changing every setting at once.
1. Get the local setup ready
Open Studio’s image workspace and choose a supported image model. Finish any required model or engine downloads and check that the selected setup is ready before starting a generation.
Use a model your Mac can handle comfortably. The model file, working memory and generation workload all matter. If you are unsure where to begin, read Which AI model is right for your Mac?
Features vary by model. Reference photos, image shapes and other controls may be available for one setup and unavailable for another. Follow the capabilities shown for your selected model.
2. Write the image you want to see
A useful starting prompt describes four things: subject, setting, lighting and visual style. You can write naturally without collecting a long list of technical keywords.
TRY THIS PROMPT
A photograph of a cobalt-blue ceramic teapot on a warm wooden table beside a window. Soft morning light from the left, a simple neutral background, visible ceramic texture and a calm, everyday feel.
That gives the model a subject, somewhere to put it and a visual direction. Start with a simple composition so you can judge whether those choices came through.
3. Discuss the idea, then use the finished prompt
In Naldar chat, you can ask for alternatives, discuss a mood or explain what you dislike about a draft. For example: “Help me make this feel more like a candid photograph and less like an advertisement.”
When you are ready to generate, put the finished image instructions in the generation prompt field. Include the details the image model needs, rather than assuming it has the entire conversation you just had.
Choose an available image shape and generate one result. Let that job finish before starting several more; your Mac is doing the processing.
4. Use a reference when it helps
If the selected model and workflow support reference photos, add an image to give the generation a visual starting point. Use a clear reference that shows the details you care about, and describe which parts should remain or change.
A reference can guide the result, but it does not guarantee an exact copy or perfect identity consistency. Repeated subjects and styles may call for a different workflow, such as a compatible trained LoRA.
5. Review before you change everything
Look at the whole image first. Did it capture the subject and mood? Then inspect details that matter to your purpose: the shape of the object, small text, hands, reflections or anything that appears inconsistent.
For the teapot example, a useful next prompt might keep the subject and setting while asking for softer shadows. Another might change only the camera angle. Save the result and its prompt so you can compare your attempts.
Keep the first experiment small
Higher resolution and heavier settings can take more time and memory. Begin with the model’s supported starting settings, then explore once you have a result you understand.
Local generation keeps the processing on your Mac. Files you share, upload or save into a synced folder can still leave the device. The privacy guide explains those boundaries.
Try a second image with one deliberate change. That comparison will tell you more about your chosen model than a much longer prompt with no clear goal.