We Stopped Sourcing Stock Photos for Storyboards. We Generate Them Instead.
When we built the storyboard section for a fire department recruitment video proposal, the first version used stock photos from Pexels and Unsplash. They were fine. Firefighters, gear, fire. But they were generic in the way stock photos always are: someone else's composition, someone else's lighting, someone else's framing. Scene 3 was supposed to be a staged training burn at night. The stock photo we used showed a daylight drill. Scene 5 was about the brotherhood inside the station. We had a photo of a guy in turnout gear standing alone.
The images weren't wrong, exactly. They just didn't sell anything.
What We Switched To
We generated replacement images with Gemini, one per scene, with prompts written to match the specific visual intent of each shot in the storyboard.
The six scenes in the proposal were: The Alarm, The Response, The Work, Beyond the Fire, The Brotherhood, and The Call to Action. Each has a timestamp, a shot list, and an emotional tone we'd already written into the proposal copy. Feeding that context into a prompt gave us images that actually reflected what we were describing.
Scene 3 came back as a nighttime exterior with orange fire light and smoke. Scene 5 showed a group of firefighters in a station bay, relaxed, in the middle of the kind of moment a recruitment video is meant to capture. The hero image had the cinematic, dark-sky quality we'd been aiming for with the parallax animation.
None of these images are what the final video will look like. That's not the point. The point is that a client reviewing the proposal can look at Scene 2 and understand: this is what "the rig rolling code through town at dusk" means as a visual. Stock photos make you explain the gap between the image and the concept. Generated images can close it.
How We Prompt for This
The approach is straightforward. Take the scene description you've already written, the tone, the key visual elements, and the timing note, then turn that into a prompt. Specify the mood, the light quality, whether it's tight or wide, whether it's action or atmosphere. If the scene is cinematic, say so. If it's supposed to feel documentary, say that instead.
You're not generating final deliverables. You're generating reference frames. The standard for "good enough" is: does this help the client visualize the scene better than the stock photo would? Usually the answer is yes on the first or second try.
Why This Matters for Proposals
A proposal page for a creative project is selling a vision before anything exists. The copy describes the concept. The pricing justifies the budget. The images are the only way to make it feel real.
When we had stock photos, the storyboard section read as a list with decorations. After switching to generated images, it read as a storyboard. The visual and the description were pointing at the same thing. That coherence is what makes a proposal feel considered rather than assembled.
The generation step adds maybe thirty minutes to the proposal build. The images live as local files alongside the proposal HTML. No external dependencies, no licensing ambiguity for internal mockup use.
For any project where the deliverable is visual, this is now the default. It's a small change that makes the proposal work harder.