Keep a character's face consistent across shots using a Flow reference image

Write a Veo 3.1 prompt intended to be used together with an uploaded reference image in Flow's ingredients-to-video mode, where the reference image — not the text description — is what should govern the character's actual facial appearance across every generated shot. The discipline here is describing action, setting, and camera in full detail while deliberately not re-describing the face, hair, or body type the reference image already establishes, since a text description that conflicts even slightly with the reference image gives the model two different sources of truth about what the character looks like, and it will not reliably favor the image over the text.

REFERENCE IMAGE
a headshot of a man in his 40s with short greying hair, uploaded as the character anchor — a note describing what the uploaded image actually shows, for your own tracking, not a substitute for the image itself; the actual appearance comes from the file, not from this line.

WHAT TO DESCRIBE IN TEXT
the man, now wearing a navy suit jacket over the same build shown in the reference, walking into a glass-walled office lobby. Describe what the character is doing, wearing in this specific shot (if different from the reference image — a costume change is fine to describe in text), and where they are, in full detail — this is exactly the information the reference image cannot supply on its own.

WHAT NOT TO RE-DESCRIBE IN TEXT
Do not restate the character's face, hair color, hair style, body type, or any other physical trait already visible in the reference image, even in passing — every one of those restated details is a chance for the text to say something subtly different from the image, and Flow does not reliably resolve that conflict in the image's favor.

CAMERA
a slow tracking shot following him from a three-quarter angle as he crosses the lobby for this specific shot.

LIGHTING AND ENVIRONMENT
bright, even daylight through floor-to-ceiling glass, a modern corporate lobby, matched to whatever look this shot needs, independent of the reference image's own lighting, since the reference image is only anchoring appearance, not the lighting or setting of the new shot.

STYLE AND MOOD
clean corporate commercial look, neutral color grade.

CONSISTENCY CHECK ACROSS MULTIPLE SHOTS
a second shot of him sitting down at a boardroom table, and a third of him shaking hands with a colleague. For each additional shot using the same reference image, repeat this same discipline — describe the new action and setting, never the face — so the character reads as the same person across every shot for the reasons stated above, not because the text happened to describe them identically each time.

WHAT TO AVOID
Do not swap in a different reference image mid-sequence for the same character without expecting a visible discontinuity — a character's appearance is only as consistent as the single reference image anchoring it, and two different photos of even the same real person can anchor subtly different appearances across a sequence.

OUTPUT
The finished prompt for this shot, followed by one line confirming no facial or body description was restated in the text, and a running list of which reference image is anchoring which character across the full sequence.