Make your own visuals!
You can now make your own visuals with AI VJ! The colorful paint brush on a film roll icon on the side-panel will open up the AI VJ Creative Director overlay.
Here you can enter a brief prompt, and improve it by adding a Style and Camera Angle prompt, and by enabling some enhancement keywords.
Then you will create the first frame of the video. This allows you to see exactly what the broad mood, style, and atmosphere of the visual is going to be. Once you are happy, you can generate a full-length visual based on this!
Also notice that these visuals are specifically tailored to be optimal for AI VJ at 60 frames per second and full HD! They will look good no matter the BPM. Just like our DLC visuals.
In fact they likely will look even a tad better than our DLC visuals, as these are made with the latest commercial models! We're looking forward to seeing what you come up with!
Notice that you can also add your own pictures as the first frame, so animated existing logos, event pictures, and other specials, are now easy to make for you!
Brewing for a while
When we first released AI VJ two years ago, being able to make ones own visuals was already listed as one of the major milestones on our road to a version 1.0 release.
The AI world was very different back then. Commercial video models weren't a thing. Usable video models themselves weren't even! The first visuals we released were created by hacking open source image models and scripts, so they would create frames affected by the previous, that we then manually combined into a video!
At the time all this still would run on consumer grade hardware. So at the time I imagined packaging whole AI models, with our pipeline, and selling that as DLC at some point, to allow you to make your own visuals without requiring knowledge of Linux, scripts and the empirically established optimal parameter tweaks of the day.
As work on AI VJ continued, new features were added, and bugs were fixed, I revisited this approach from time to time.
However, packaging everything so it would reliably run on WIndows, on the wide variety of hardware and systems out there, was a nightmare. On our Linux production machines even a driver update, or starting another program that ate up some VRAM, could sometimes break things; this brittle-ness would be magnitudes worse across user systems.
Another shift that has been happening is that video models were becoming ever more demanding. While the first models we used would (with hacks and squeezing optimizations) run on mid-range hardware, current top-tier open weight video models don't. They don't even run on the top-tier hardware owned by a small fraction of the Steam user base, any more.
Hence we decided to implement this feature with the use of commercial models. So anyone, regardless of their hardware, can enjoy the power of AI!
Commercial models
Commercial models cost money. Just like hardware (a H100 is over $27,000), electrical power, and food. It is what it is with the laws of physics and capitalism as they are.
The current, alpha-series of Creative Director visuals provides full HD, 60fps videos at a cost of $ 1.26; a little over a dollar each, less with a 10% distribution rights discount. With images being only 8 cents.
Our aim is to always offer the best, at a resolution and fps suitable for show time!
Once newer models become available, and we can make a reseller deal with them for access, we will release a next series (bravo series), and adjust the price of the alpha series accordingly.
To celebrate the release of Creative Director, new users of Creative Director will get $ 2 free credit. No purchase required. Enjoy!
Bugs Fixed
Several bugs were fixed, and more small improvements made:
- Fixed a bug that caused the play-time for the jump back play mode to be shorter than it needed to be. Now the full length of the visuals is used
- Several small fixes
If you reported (other) bugs, we're keeping a list and working through them! Please keep reports coming! We won't know if you don't tell us (best place to do so is in our Discord)
Join us on Discord!
Feature requests & bug reports are welcome. Either here or on our Discord!