Midjourney v8.1

Image Models

imagegeneratetext to imageimage to imageartcreativestylereference

Overview

Midjourney v8.1 on Youchuan: the latest full upgrade with much stronger prompt adherence, richer detail, and native 2K HD rendering (--hd outputs four high-res images per task, ~5x faster than the previous generation). Text-to-image and image-prompted generation. Two reference channels: Image Prompt inputs (reference images that influence content/composition/color, weighted by --iw) and Style Reference inputs (--sref, transfers only aesthetic/style, weighted by --sw). Quote text in the prompt for accurate text rendering. Note: v8.1 does not support character/omni reference; for subject consistency use an image prompt with a higher --iw plus a precise text description.

Optional inputs: up to 4 image prompts (content/composition reference) and up to 4 style references.

Image
Style Ref 1
Style Ref 2
Style Ref 3
Style Ref 4