šø Using the 'describe' feature in Discord, you can ask mid-journey to analyze an image and provide prompts based on what it sees.
š” After uploading an image, mid-journey generates four different prompts related to the image's content, such as robotic person, futuristic makeup, or cyberpunk artwork.
š The prompts provided by mid-journey can be selected and submitted for further analysis and exploration.
šø Adding the original reference picture as an image prompt is crucial for accuracy.
š By copying and pasting the image address, the reference picture can be used effectively in the prompts.
āļø Using the reference picture significantly improves the accuracy and quality of the midjourney prompts.
š By adding a weight to the reference picture, you can control its importance in generating the prompts.
š· Increasing the weight value results in more prompts that closely resemble the reference picture.
āļø There are additional techniques available to further improve the generated prompts.
š Adjust the stylize or chaos value to make Midjourney follow the prompt closely or have creative freedom.
š” Lower stylized value for a more literal interpretation of the prompt.
šø Adding chaos value changes the variety of pictures generated.
š· The four pictures in the grid exhibit a lot of variety and are highly different from each other, especially at chaos 40.
š° The speaker announces a special offer without relying on sponsors, offering a prompt pack with 51 unique prompts and 69 example images for sale on their website.
š An additional feature allows users to re-roll the describe prompts to generate four more prompts.
š The describe prompt in Midjourney can sometimes mention an artist with a hyperlink or a name without a hyperlink.
šØ Midjourney has the ability to interpret made-up names and may round aspect ratios to the nearest 32-pixel value.
š” Please keep in mind that describe uses a model that is not native to Midjourney to describe pictures.
š Mid-journey prompts can accurately describe something but may not always be able to recreate it exactly.
š¼ļø The example of using describe on a picture with text shows that mid-journey may not generate the text accurately.
š„ There is a video on how to maximize the benefits of mid-journey's describe feature.