Transit wants a short explainer video built from a script and approved station photos. Which generation approach should the team use?

Transit that wants a short explainer video built from a script and approved station photos should generate video from the script prompt using those station photos as reference media. The script supplies narrative; approved stills keep branding and location look consistent. Reference-conditioned video generation produces an on-brand explainer rather than consuming the station photos as style fine-tuning data, which retrains model style instead of conditioning this generation on approved look. That is the generation approach the stem describes. Generating video from the station photos as image prompt using the script as style guidance fails by concept: it inverts the required inputs, letting stills drive the video while reducing the script narrative to stylistic hints instead of using the script as the prompt with photos as reference media. Generating video from the script prompt using the station photos as style fine-tuning data fails because it misuses approved references for model training rather than for reference-conditioned generation, adding training overhead and rights concerns without meeting the script-plus-reference explainer need. Generating video from the script prompt using the station photos as start frame images fails because start frames only fix the opening composition, not sustained generative references, and do not condition the video on approved branding and location look throughout. Exam caveat: script-plus-approved-reference video generation is the civic pattern—not photos-as-image-prompt with script-as-style-guidance, script prompt with photos as fine-tuning data, or script prompt with photos as start frame images. Operational check: assemble the approved script and station photo references, generate the explainer with those inputs, review for brand and rights compliance, and publish only after transit communications signs off.