Master Midjourney: Precise Image Control with /describe, Seed & Image Weight

1. Using the /describe command to extract prompts

Open Midjourney, type /describe in the chat, and upload a reference image. The AI returns four descriptive prompts; choose the one that fits your needs and feed it to the /image command to generate a new picture.

2. Combining reference images with prompts for precise control

Drag the reference image into the chat, right‑click to copy its link, then use /image with the format {image‑link} {prompt}. The AI will generate a new artwork that closely follows the reference.

3. Advanced parameter – image weight (--iw)

The --iw flag controls how much influence the reference image has compared to the textual prompt. In Midjourney V5 the valid range is 0.5‑2; a higher value (e.g., --iw 2) makes the result more similar to the reference.

4. Determining the seed value of an image

Send the generated image to your Midjourney inbox (email envelope). The service replies with the seed number, which can be used for further fine‑tuning.

5. Using the seed for micro‑adjustments

Copy the original prompt, add any new descriptive words, and append --seed XXX (replace XXX with the obtained seed). This reproduces the image while applying the modifications, such as changing eye color.

6. Changing a character’s clothing with Midjourney V5

Generate a model photo using a detailed prompt (e.g., "A young beautiful Chinese model, full body, high perspective, smiling, clean background --ar 9:16 --v 5"), then overlay the brand’s clothing in Photoshop. Upload the combined image, copy its link, and run /image with a high --iw (e.g., --iw 2) to produce a realistic, brand‑styled illustration.

7. Useful keyword examples for model generation

She is a chef, Korean, youthful, arms folded, light gray background, soft, happy

She is a Chinese sales assistant, youthful, thumbs up, baby blue background, soft illumination, happy

8. Summary of key points

1) Generate the desired style : Use reference images, combine with prompts, and adjust --iw to steer the AI toward the target aesthetic.

2) Fine‑tune existing images : Retrieve the seed and re‑run the prompt with --seed for precise modifications.

3) Brand clothing replacement : Overlay clothing in Photoshop, then feed the composite back to Midjourney with a high image weight.

AI image generation is becoming a powerful tool across industries; sharing techniques and collaborating can accelerate its development and practical adoption.