Midjourney V8.2 is now on XinYu AI — pick it from the model list in any image node. Midjourney describes V8.2 around two things: images that are bolder and more creative, and a model that learns your personal taste better. We built the style tools Midjourney users rely on most around exactly that.
Bolder images
These four were generated on XinYu with V8.2 in HD: a fashion editorial, a film still, an anime key visual and architecture.
No reference image for text-to-image, 1 for image-to-image, 2–5 to blend. Each run returns a set (usually 4 images), stacked in one node — open it and pick your favorite.
Step by step into your style
A new Style button sits next to the aspect ratio. It brings your taste in three ways: your Midjourney personalization (P code), a moodboard code, and a style reference image.
Same prompt, same seed — left to right with P code, moodboard and style reference switched on in turn:
Stylize, chaos and weird stay in the aspect-ratio panel, grouped the way Midjourney groups them.
Style presets: your taste in one click
Save a combination of P codes / moodboard / style reference / values as a preset — name it, add 0–4 thumbnails, and apply it from the Style button. Presets live in your account across devices; after tweaking you can Update or Save as new, and empty thumbnail slots fill in with images made from that preset.
Style transfer: same composition, any style
Select any image on the canvas → More → Style transfer, describe a look (style, material or medium — the content comes from your image) or tap a common one. The composition and pose stay; only the style changes. Four results per run:
Pricing
Billed per run, one set per run: 11 credits standard, 17 HD; style transfer 17.
CLI and MCP
Update @xinyuai/cli and @xinyuai/mcp-server to 0.1.40 to use --moodboard, --sref, --mj-preset <name> and --mj-style-transfer (same-named parameters in MCP).