Input images + output examples
As I have been working with this model, I have learned that it isn't necessarily magic and you don't get the amazing results from the documentation without a lot of hard work.
There are a lot of factors:
The size of the init_image. You could possibly resize your input images to yield interesting results.
The init_image itself. There are better or worse ones.
The various configuration knobs, like guidance_scale, controlnet_conditioning_scale, strength, and num_inference_steps.
It would be really great if your examples could be expanded to show the raw input images (qr code + style image ), the prompt, and the settings used, to generate the output image. Your example code uses images which are in S3, which are unavailable for the public to view.
In the same vain, it seems to me like this is best for blending the two input images together, I have yet to see how the prompt really improves things, so potentially examples could help there.