Back to library
Stable Diffusion#marketingStandard promptSaves ~15min

Photorealistic stock-style image to replace generic stock photography

Photorealistic stock-style images that don't look AI-generated. Covers which models to pick, how to word the prompt so the camera feels real, and a quick checklist to spot AI tells before you publish.

rach_maeve4 June 2026
0.0(0)

Fill in the Template

0 / 7 filled

Your Built Prompt

Updates live as you type
I need a photorealistic stock-style image for a marketing piece and the free stock libraries don't have what I need (or every founder is using the same shot). I want to generate one that looks like real photography, not AI.

What the image is for: {{use_case}}
Scene I need: {{scene_description}}
People in the frame (or none): {{people_in_frame}}
Lighting / time of day: {{lighting}}
Camera feel I want (DSLR, film, phone, editorial): {{camera_feel}}
Things that scream "AI image" and must NOT appear: {{ai_tells}}
Resolution: {{resolution}}

Give me a Stable Diffusion workflow that produces a photo, not an AI render:

1. CHECKPOINT - recommend a photoreal checkpoint (Realistic Vision, epiCRealism, Juggernaut XL). Give me 2 options + when to use each.

2. POSITIVE PROMPT - comma-separated, ordered by importance:
   - Camera tokens FIRST: 35mm, f/1.8, shallow depth of field, natural light (or whatever matches {{camera_feel}})
   - Scene: 10-15 tokens describing {{scene_description}}
   - Lighting tokens: match {{lighting}}
   - Skin / surface realism tokens if people are in the frame: visible pores, natural skin texture, candid expression
   - Imperfection tokens: slight grain, real photo, not staged

3. NEGATIVE PROMPT - the "AI tell" killer. Bake in {{ai_tells}} plus the universal SD tells: plastic skin, overly smooth, glossy, oversaturated, cartoon, 3D render, illustration, AI generated, deformed hands, asymmetrical face, text artefacts, watermark.

4. SAMPLER + CFG - recommended for photoreal. (Hint: lower CFG than you'd expect; high CFG makes images look "AI-cooked".)

5. RESOLUTION + UPSCALE - generate at native checkpoint resolution, then upscale to {{resolution}} with a photo-realistic upscaler.

6. FAIL TELLS - the 5 things to scan the output for before approving: hands, ears, teeth, reflections, edges of small objects. If any look wrong, regenerate.

7. ETHICS - one-line reminder that if real-looking people are in the frame, the brand should disclose AI generation per AU advertising standards.

AU English. Write so a marketer who has never used SD before can follow.
Run in

Fill in the required fields above to copy or run this prompt.

Tags

Community Feedback

0 comments
Be the first to leave feedback on this prompt.

Related Prompts

More in Marketing