AI Speed Box
Photoshop (Beta) New Feature: Select Body Parts
- AI Image
- Adobe Photoshop
2024/12/19
Available Now (Available)
Report: AIBRAHAM
Midjourney Office Hours (2024-12-04)
- AI Image
- Midjourney
V7 is scheduled to be released in January 2025. Video release is expected to be delayed. Storytelling is also delayed. Moodboard: Personalization to be applied as a set of images.
https://x.com/blackowl777/status/1864496878114558025
2024/12/04
Coming Soon
Midjourney Office Hours (2024-12-04)
Category
- AI Image
Gen
- Midjourney
Date
2024/12/04
Summary 🍀🧸
V7 is scheduled to be released in January 2025. Video release is expected to be delayed. Storytelling is also delayed. Moodboard: Personalization to be applied as a set of images.
URL
https://x.com/blackowl777/status/1864496878114558025
Release
Coming Soon
Midjourney Office Hours (2024-12-04)
Original
Midjourney Office Hours 2024-12-04
(Source: GenIArt)
(Source: GenIArt)
V7 Model Status
January 2025 Release Confirmed: December release no longer possible
Model training progressing well
Team working on optimizing text dataset balance
Challenge: More text can improve quality but may degrade visual output
V7.1 and subsequent versions in development
Will introduce new aesthetic and personalization systems
Planning slower but higher quality model variants
Similar to Q2/Q4 approach
Responding to user requests for higher quality options
Video Model Strategy
Key Decision Point: Affordable vs. Premium Model
Community poll showed 50/50 split on preferences
Three potential approaches under consideration:
Affordable model with lower performance
Premium model with highest possible quality
Limited-scope model with balanced cost/performance
Business Constraints
Current video models require significant investment ($1B+ range)
ROI challenges due to Midjourney's self-funded business model
Technology limitations make affordable, high-quality video currently unfeasible
Plan user ranking collection for video model training
Upcoming Releases
Personalization Profiles
Multiple personalization support
New "mood board" feature: personalization from a bunch of images
Storytelling Tools
First experimental version
Demo planned for next office hours
Potential capacity issues expected at launch
Future Development
Personalization Improvements
Enhanced character references for V7
New “omni reference” system in development
Will allow specific object/design integration
Big batch release (December vs. V7 integration)
Final video model approach selection
Release schedule and prioritization for 2025
Architectural Changes
Moving toward modular architecture
Separating personalization from model versions
Decoupling datasets from model versions
Independent scaling and architecture modifications
Translation
Midjourney Office Hours 2024-12-04
(Source: GenIArt)
(Source: GenIArt)
V7 Model Status
Confirmed for release in January 2025 : December release is not possible
Model training is progressing well
Team working to optimize text dataset balance
Challenge: Using more text can improve quality, but may result in poor visual output
V7.1 and subsequent versions in development
New aesthetic and personalization systems to be introduced
Planning a slower but higher quality model transformation
Similar to Q2/Q4 approach
Providing high quality options according to user requests
Video Model Strategy
Key Decision Points: Budget vs. Premium
In a community poll, the preference was split 50/50.
Three potential approaches are being considered:
Low-cost model with low performance
Premium model with the best possible quality
Limited scope model that balances cost and performance
Business constraints
Current video models require significant investment (in the $1 billion+ range)
Midjourney's self-funded business model leaves ROI challenges
Due to technological limitations, low-cost, high-quality video models are currently unfeasible.
User Rank Collection Scheme for Video Model Training
Coming soon
Personalized profile
Supports multi-personalization
New “Mood Board” feature: Personalization from a collection of images
Storytelling tools
First experimental version
Plan a demo during the next office hours
Potential capacity issues expected at launch
Future Development
Improved personalization
Enhanced character reference for V7
A new “omni-reference” system is under development
Ability to integrate specific objects/designs
Massive Batch Release (December vs V7 Integration)
Selecting the final video model approach
2025 Release Schedule and Priority Setting
Structural change
Moving to a modular architecture
Separate model versions and personalization
Separate datasets from model versions
Independent extensions and architectural modifications
