The quality of an AI-generated image depends more on the prompt than on the model itself. With the right structure, even free tools can produce surprisingly photorealistic results. With a vague prompt, even the most expensive model will deliver something generic.

The prompts below were structured and tested to work well in Midjourney v6, Stable Diffusion XL, and DALL-E 3. Adapt the main subject and keep the structure.

The structure of a good realistic prompt

Every effective prompt for realistic images follows a hierarchy of information: subject + context + lighting + camera/lens + technical finish. Each layer adds specificity and reduces the model's "guesswork" — which, paradoxically, produces more convincing results.

💡 Tip: For Midjourney, add --ar 3:2 --v 6 to the end of any prompt for a standard photographic ratio. For Stable Diffusion, prefer models like Realistic Vision or DreamShaper for more photorealistic results.

The 10 prompts

1. Female portrait with natural light

PROMPT 01
Portrait of a 30-year-old Brazilian woman, natural window light from the left, shallow depth of field, Sony A7IV 85mm f/1.8, soft skin texture, editorial photography style, muted warm tones

The secret is specifying the camera and lens. This forces the model to simulate the real optical characteristics of each piece of gear — the compression of an 85mm, the aperture that creates bokeh, Sony's characteristic color.

2. Product in a minimalist studio

PROMPT 02
Commercial product photography, dark glass perfume bottle on a white marble surface, studio lighting with soft box, reflection on surface, isolated on white background, high-end luxury brand aesthetic, 8K detail

Ideal for e-commerce. Mentioning luxury brand aesthetic and 8K detail steers the model toward a premium finish.

3. Night urban architecture

PROMPT 03
Night street photography, São Paulo downtown, neon reflections on wet pavement, long exposure effect, Fujifilm X-T5, cinematic grain, blue and orange color palette, documentary style

4. Ultra-detailed macro nature

PROMPT 04
Macro photography of a dewdrop on a green leaf, morning light, extreme close-up, Canon MP-E 65mm macro lens, water droplet refracting the background forest, vivid natural colors, National Geographic style

5. Food for social media

PROMPT 05
Flat lay food photography, açaí bowl with granola and tropical fruits on a dark wooden table, overhead shot, natural soft light, Instagram aesthetic, earthy tones, professional food stylist composition

6. Sunset landscape

PROMPT 06
Golden hour landscape, Chapada dos Veadeiros, Brazil, vast cerrado savanna, dramatic sky with orange and purple clouds, wide angle 16mm lens, high dynamic range, photojournalism quality

7. Editorial male portrait

PROMPT 07
Editorial portrait of a 45-year-old man, salt and pepper beard, confident expression, dark studio background, dramatic Rembrandt lighting, Hasselblad medium format, sharp focus on eyes, fashion magazine cover quality

8. Modern office interior

PROMPT 08
Architectural interior photography, modern coworking space, floor-to-ceiling windows, morning light, plants and wood accents, wide angle tilt-shift lens, real estate photography style, clean and airy atmosphere

9. Wildlife in its natural habitat

PROMPT 09
Wildlife photography, jaguar resting on a tree branch in the Amazon rainforest, dappled sunlight through leaves, telephoto 500mm lens, shallow depth of field, BBC Earth documentary quality, golden eyes sharp focus

10. Documentary street scene

PROMPT 10
Street documentary photography, Feira de São Cristóvão, Rio de Janeiro, candid moment, elderly vendor selling typical food, black and white with slight grain, Leica Q2 28mm, Henri Cartier-Bresson compositional style

💡 Tip: After generating the image, use ImageTools' Image Compressor to optimize the file for the web without losing the visual quality of the result.

Modifiers that always improve realism

Some terms act as quality amplifiers in almost any prompt. Add them as needed: shot on film, photorealistic, hyperdetailed, professional photography, award-winning photo, 8K resolution, cinematic lighting.

Avoid generic terms like "beautiful" or "amazing" — they don't add any technical information for the model. Instead, describe what you want to see with visual precision.

Your generated image deserves the best treatment

Convert, compress, or resize AI-generated images with ImageTools' free tools.

Convert image now

Frequently asked questions

Do these prompts work in Portuguese?
For more consistent results, use English — the main models (Midjourney, Stable Diffusion, DALL-E) were trained mostly on English descriptions. DALL-E 3 handles Portuguese reasonably well, but English still tends to give you more technical control.
Why mention specific cameras and lenses in the prompt?
The models learned to associate camera brands and models with specific visual traits: Fujifilm's color, Leica's sharpness, Canon's bokeh. The camera doesn't need to actually appear in the photo — the model uses that information as a visual style reference.
Which free AI tool gives the best results with these prompts?
Stable Diffusion XL with the Realistic Vision model (available in Automatic1111 or Forge) is the free option with the best photorealistic results. For online use with no installation, Adobe Firefly (free with an Adobe account) and Ideogram are good no-cost options.