September 2026
Reference images: the part most people skip
The prompt gets the attention. The reference images do the work. Here's how I think about them after building the RESILIENT character sheet and object set.
The character sheet fixes the face
A character reference sheet built from my own photos — face, outfit, accessories — is what kept the character consistent across every shot in RESILIENT.
No amount of prompt detail about 'the same person' worked as well as one reference image. The model trusts an image more than a description.
Object references pin the product
For the phone, sunglasses, coffee cup, and pocket logo I used separate object references. Each one told the model exactly which object to place and where, so the scene read correctly.
Match each reference image to the corresponding line in the prompt. A reference without a matching description drifts; a description without a reference blurs.
Choose references for clarity, not beauty
I pick the clearest image, not the prettiest. A flat, well-lit shot of the object tells the model more than a dramatic one with shadows across the key detail.
For the character, I chose photos where the face and outfit were fully visible — not the most flattering angle, the most informative one.
Location references are the next layer
After the character and objects, the location reference is what makes a shot feel placed rather than floating. This is the part I'd strengthen next time — more location references, fewer location adjectives in the prompt.
Keep reading
- See the references in the RESILIENT tutorial
- My AI tools tutorial
- How I write a video prompt that actually holds together
- Behind the RESILIENT video: what the tutorial doesn't show
- The AI tools I'm testing this month
- What I learned shipping AI content publicly
- My creator routine in Dubai