Consistent AI-Generated People Images
Budget: £5 – £10 GBP
This project centres on producing photorealistic image sets for AI research. For every assigned subject I’ll need 4–6 high-resolution shots that keep the exact same face and hairstyle while placing the person in convincingly real environments—think natural street scenes, home interiors, offices, cafés, parks and similar everyday backdrops. Lighting, shadows and colour grading must match the chosen setting so the final result reads as genuine photography rather than a composite.
Over the next few months (and potentially up to a year) I’ll supply briefs in batches. Each batch will specify the demographic focus—ethnicity, age group or other traits we decide on together—to ensure we cover the diversity the research requires. Consistency across each mini-set is critical: the face must remain identical from frame to frame, with no drifting features or hairstyle changes, yet each image should offer a fresh angle, pose or context so the data isn’t repetitive.
Use whichever tools you prefer—Stable Diffusion, Midjourney, DALL-E 3, ControlNet, Photoshop or your own specialised pipeline—as long as they achieve the realism and continuity described. Please retain layered or prompt files so adjustments can be made if small tweaks are needed later.
Deliverables per subject
• 4–6 JPG or PNG images, minimum 3000 px on the longest side
• Identical facial features and hairstyle across the set
• Authentic, artefact-free real-world backgrounds
• Source prompts or editable files (PSD, .blend, etc.) for reproducibility
I’ll review the first set as a proof of concept. Once quality and style are locked, we’ll move into a steady workflow with regular milestones and feedback.
Over the next few months (and potentially up to a year) I’ll supply briefs in batches. Each batch will specify the demographic focus—ethnicity, age group or other traits we decide on together—to ensure we cover the diversity the research requires. Consistency across each mini-set is critical: the face must remain identical from frame to frame, with no drifting features or hairstyle changes, yet each image should offer a fresh angle, pose or context so the data isn’t repetitive.
Use whichever tools you prefer—Stable Diffusion, Midjourney, DALL-E 3, ControlNet, Photoshop or your own specialised pipeline—as long as they achieve the realism and continuity described. Please retain layered or prompt files so adjustments can be made if small tweaks are needed later.
Deliverables per subject
• 4–6 JPG or PNG images, minimum 3000 px on the longest side
• Identical facial features and hairstyle across the set
• Authentic, artefact-free real-world backgrounds
• Source prompts or editable files (PSD, .blend, etc.) for reproducibility
I’ll review the first set as a proof of concept. Once quality and style are locked, we’ll move into a steady workflow with regular milestones and feedback.