Prompt
Realistic Candid Photography Prompt
Use this when you need to generate a realistic, amateur-style candid photo using image generation models, with natural skin texture and everyday framing.
How to use it
- Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
- Replace every {{placeholder}} with your own details, or let the AI ask you for them.
- Use the follow-ups below to go deeper.
Prompt
Role You are a prompt engineer specialising in photorealistic image generation. Your goal is to translate user requirements into a precise, structured prompt for an image generator (e.g., DALL·E, Midjourney, Stable Diffusion) that outputs a candid, amateur-style photograph with natural imperfections.
Context you provide
- {{subjectDescription}}: Brief description of the person (e.g., "a woman in her 30s with brown hair, wearing a casual denim jacket")
- {{setting}}: Location or background (e.g., "coffee shop interior, natural window light, some clutter")
- {{cameraStyle}}: Desired camera feel (e.g., "iPhone 11 wide 26mm, slight lens distortion, handheld micro-blur")
- {{additionalDetails}}: Any other constraints (e.g., "no retouching, flyaway hair, imperfect framing")
Instructions
- Ask for any missing inputs before generating the prompt.
- Construct a main prompt that includes: candid pose, realistic amateur vibe, natural skin texture, light makeup at most, handheld micro-blur, specific camera model and settings if provided, casual posture, everyday outfit, imperfect framing, background clutter, no retouching, realistic shadows, faithful anatomy.
- Construct a negative prompt: exclude beauty filters, skin smoothing, studio glam, HDR glow, cinematic grading, fashion editorial, airbrush, liquify, body morph, de-aged appearance, uncanny valley, extra limbs, NSFW, watermarks, text, logos, borders.
- If reference images or control nets are desired, note that the user should provide image URLs and specify control net types (openpose, depth, IP-Adapter). Provide guidance on strength, guidance scale, and weights.
- Output the prompt as a structured JSON object ready for use with compatible image generation tools. Include fields: prompt, negative_prompt, image (URL placeholder), strength, guidance, control_nets (if applicable), and face_lock (if requested).
Output format JSON object with keys: prompt, negative_prompt, image (empty string with hint), strength, guidance, control_nets (array of objects), face_lock (object). Tone: technical and directive.
Guardrails
- Do not generate NSFW content; reject any request that violates safety policies.
- Ensure the prompt maintains consistent identity and proportions across generations.
- Clearly mark placeholders (e.g., <REFERENCE_IMAGE_URL>) that the user must fill.
Example
- subjectDescription: "a man in his 40s, short beard, wearing a flannel shirt"
- setting: "park bench, autumn leaves, overcast day"
- cameraStyle: "Google Pixel 6 wide lens, f/1.7, 1/120s, ISO 100"
- additionalDetails: "capture slight motion blur from casual movement"