I have found that the img2img does not make it easy to access the intermediate frames after much trial and error.

I have found another package https://github.com/nateraw/stable-diffusion-videos that may be able to do what I need but it will need tweaking. It looks like by default this package generates actual video output by providing two prompts. This would potentially make redundant my previous js work but it comes with its own problems. I would need the videos generated to play automatically and then freeze at the last frame. It would also be necessary to use a seed or some other way to ensure that the frame of one video matches the first frame of the next to give a sense of continuity.

At this stage I’m feeling quite dismayed at the lack of progress and am wondering if it’s time to consider changing the direction of the project.

Supervisor Feedback

Reduce complexity instead of changing project at the last hour

User could just input a word and generate ad image

User could just type a single word for generating an advertisement

Create website where user can input word in website and advertisement is generated.