Preface

I was asked if I can test the SD Prompt Reader node as I did it for the Load Image node [2]. Here are the results in short. After a first run, I reorganized everything again to make sure that I was using the same initial conditions, the same nodes and the same settings.

Workflow Preview

Figure 1: Workflow preview

Comparison of Input and Output

The comparison of input and output is:

Figure 2: Visual comparison of input and output

Comparison of the Comparison Results

The next image shows the direct comparison of old and new result. The latter result is at first glance more worst than the first result.

Figure 3: Old and new result w.r.t the comparisons

The comparison without using a shadow can be seen in Figure 4.

Figure 4: Old and new result compared w.r.t the comparisons

There is no difference. One can state that Load Image node and SD Prompt Reader node behave the same way.

Conclusion

To be sure that what I say and do is absolutely correct, the results from the articles must be thoroughly checked again. I checked the Python code and I checked my results twice. Both loaders using the same saver will result the same difference.

References

[2] https://civitai.com/articles/13611/comfyui-loading-and-saving-images

[3] https://civitai.com/articles/13673/load-and-save-a-jpeg-image-in-python

[4] https://www.copus.io/work/407de638f2418eec5fd2baf9ad58d2d2