High level Midjourney overview notes from todays stream:
### Platform Overview & Navigation
- Midjourney web platform recommended over Discord
- Key sections:
- Explore page: Browse community generations, search by keywords or similar images
- Create page: Main generation interface
- Organize page: Gallery management and folder system
- Personalize page: Customize AI preferences
### Personalization System
- Two types of profiles:
- Standard profiles: Broad aesthetic preferences across all imagery
- Moodboards: Custom collections of reference images
- Profile development:
- Minimum 40 ratings to unlock
- Optimal results at 1000+ ratings
- Global profile: Rate broadly for general aesthetic
- Specialized profiles: Rate selectively for specific styles
- Implementation:
- Add with --p parameter
- Can combine multiple profiles
- Profile codes update with new ratings
### Style References & Image Usage
- Three reference types:
- Style references (paperclip icon): Decouples subject from aesthetic
- Image references (image icon): Integrates actual image elements
- Character references (user icon): Currently for humanoid figures, expanding in v7
- Weight parameters:
- Style weight (–sw): 0-1000, default 100
- Image weight (–iw): 0-3, default 1
- Character weight (–cw): 0-100, default 100
### SREF Codes System
- 4.6 billion possible style reference codes (2^32)
- Usage:
- –sref random for exploration
- Can blend multiple codes using :: syntax
- Weight adjustment format: code::number
- Advantages over images as style references:
- More dynamic/flexible results
- Better for blending and adjusting weights
- Can be easily shared and reused
### Organization Features
- New folder system:
- Create project-specific folders
- Generate directly into folders
- Bulk selection and management
- Save searches (formerly smart folders):
- Automatic collection based on keywords
- Dynamic updating
- Quick navigation:
- Date sidebar for rapid browsing
- Filter by profile, likes, or upscales
### Technical Parameters
- Aspect ratios:
- Default: Square (1:1)
- Portrait: 4:5, 9:16
- Landscape: 16:9, 3:2
- Can be any whole number ratio
- Seed parameter (–seed):
- Fixes generation starting point
- Useful for comparison testing
- Random by default
- Not useful in most cases, use only as needed
### Upcoming V7 Features
- Omni reference system replacing character reference
- 3D capability:
- Light fields, not meshes
- Interactive camera movement
- Not mesh-based 3D assets
- Expected v7 release: “Very, very close” (within weeks)
at a high level, srefs converge into a single cohesive style, while moodboards maintain more diversity
for example, lets say you have 5 images, each with a unique style
when you use them as style references (srefs) in Midjourney, you’re essentially blending the styles of those different images into a single, cohesive aesthetic.
Imagine mixing multiple paint colors. All the individual colors merge into one new color. Same thing happens when you provide multiple style references. Midjourney “blends” them into a singular new style, which then gets consistently applied across all images in your output grid.
On the other hand, if you create a moodboard with those images and use the moodboard code in your prompt, instead of the images converging into one blended style, the model will generate a more diverse set of images that reflect the different styles from that collection. This results in a more diverse image grid, each image in the grid potentially leaning toward a different style within the moodboard rather than a single uniform look.
So style references (srefs):
> Blends styles of all images/codes used into one new “style color”
> produce image grid with a single cohesive style that gets applied to every img in the grid
While moodboards:
> Keep each style somewhat distinct
> Produce image grids with more diversity, influenced by different elements in the moodboard
hope that makes sense!
When IP Powerhouses Unite 🤝
We are delighted to strategically partner with @QORPOworld, the masterminds behind the flagship game titles @CitizenConflict & @playaneemate.
This trailblazing partnership & $RCADE's seamless interoperability allow for unprecedented ecosystem expansion, capable of onboarding the next billion gamers.
Together, we will eliminate barriers to entry, onboarding Web2 & Web3 gamers into our IP-rich ecosystem.
This is just the beginning of our mammoth vision to unite gamers worldwide!
I just contributed to the @AcalaNetwork Crowdloan to get my piece of the Ethereum-compatible DeFi and liquidity hub coming soon to @Polkadot. Join now to get a 5% ACA bonus when you contribute! https://t.co/qGn542O9K6