Midjourney now supports the /describe command, which works in reverse: you upload an image, and it generates a prompt for you.
What is this for?
1. Like the “rewrite text” function, but in the format of “redraw.” You upload an image, get a prompt, modify it if necessary, and receive an updated image.
2. To add multimodality to your networks while multimodal networks like GPT-4 are still unavailable. Want a linguistic model to work with the uploaded image? Describe it using /describe, and then use the image description as part of the prompt. This will already give an 80% boost compared to “bare” networks that only work with text.
P.S. In the image, we see the path: original image → prompt → generated image.
