Image/video generation workflow and vehicle edit tests

Community creator documents image/video workflow and vehicle edit tests.
Key insights
Video generation demands clear prompts and world-model consistency
The author reports that video generation requires stronger prompt skills and clearer descriptions than image generation, and emphasizes that temporal and cross‑frame consistency—what the author calls a 'world model'—is critical for accurate and coherent video output. The post notes that if a model supports both consistency and a world model, it can both recognize and remember scene elements across frames.
Workflow: image selection, Nanobanana Pro and stepwise editing
The workflow presented includes selecting a source vehicle image, sending it to Nanobanana Pro with prompts for high‑quality white-background studio shots, and stepwise editing: specify wheel style, stance, body width, large Helvetica 'JKW' decals, Red Bull stickers, red calipers and dual spoiler while keeping other areas unchanged. The author documents concrete prompts and a multi-step process for consistent edits.
Free vs paid model access and practical token use
The author notes limitations of a free 'ImDream' account with daily 80 points, which suffices to generate eight start‑end frame videos, while paid versions or owning Veo3 would improve effects and add audio. The post recommends conserving free credits and documents prompt and pitfall notes published on an external knowledge base for community reference.