Creator builds luxury fashion spot with GPT, Nano Banano 2, and Seedance 2.0
A creator documented a workflow for generating a high-end fashion spot using GPT for concept iteration, Nano Banano 2 for character design, and Seedance 2.0 for animation, then assembled the clips in After Effects.
"The iterative prompt-refinement loop with GPT is the key to controlling output quality across both still-image and video stages," a creator working on a luxury-meets-underground ad campaign explained in a recent workflow breakdown.
The project, titled "DUALÉ — BEAUTY LIVES IN BOTH," began with GPT to translate a mental concept into detailed text prompts. After generating an initial character image and refining it through multiple GPT chat iterations, the creator moved to Nano Banano 2 to finalize the character design. The Nano Banano 2 render served as the visual anchor for the rest of the project.
For animation, the creator used Seedance 2.0 — but only after spending significant time in GPT describing every detail of the desired motion and shot composition. The prompt was refined and analyzed with the language model before being fed to Seedance. That process yielded 7 to 8 animation clips ranging from 8 to 15 seconds each, which were then cut and assembled in After Effects to form the final spot.
The project frames itself as a "fake ad" exercise in the luxury/underground aesthetic space, with no brand or commercial client attached. The creator emphasizes the iterative prompt-refinement loop with GPT as the key to controlling output quality across both still-image and video stages.
