Etiquette is extremely important in these situations.
The image(s) above in this post have not been modified/iterated extensively. As such, they do not meet the minimum expression threshold, and are in the public domain.
The top three were 2.0, the last one used their "kissing" template.
The image(s) above in this post have not been modified/iterated extensively. As such, they do not meet the minimum expression threshold, and are in the public domain.
Some experiments using Vidu's reference-to-video feature with the old G1 package art as the character reference.
These were done with a minimum prompt, a standardized live-action film prompt I use for testing.
While clearly not photoreal, I kind of like the odd near-animated look that it provides. Vidu's reference feature can alter the look of the referenced item to a degree, but art style conversion is difficult for it, requiring the changes be made prior to the creation.
Doing so produces a much more live-action look, though less artsy due to it.
Process and some prompts under the fold.
I used midjourney's retexture overpainting process to generate a stack of miscolored overkills that I composited and adjusted into a quick-and-dirty reference.
The prompt I used for this:
full sized mechanical puppet-prop for live-action film, photographed against white background. A robotic dinosaur with angular, body design aesthetics, blue lower jaw, claws, legs, gunmetal body and head, silver missile launcher. 5k, photograph taken with kodak, official media, promotional image, white background/backdrop
I call it a Q&D reference because if you're doing anything more fully realized, you want front/back/side references, model sheet style. Otherwise, your details will shift if the character ever turns around, and the robot is going to be reluctant to do that in general if it doesn't have back-reference.
The backgrounds, too, were reference prompts. This is why Tulsa Gamma Ray is spelled correctly on the sign. This also helps with aesthetics, as your text prompt needs to be robust with anything not shown in an image prompt or else it will take on an uncanny or "on a greenscreen" stage look.
My standard prompt I used is:
live action motion picture film scene, 1989, filmed on 35mm film.
directed by stanley kubric and luc besson. ILM, weta digital, stan winston studios, jim henson creature shop, masterful traditional practical special effects, costuming. Professional cinematography and lighting. dramatic and cool.
live action film scene, professional lighting and cinematography, masterful practical special effects, costuming and set design. filmed in 1989
These were part of a minimal-influence test, to see what the vidu system would do with minimal direction. It does seem to be able to pick up certain context clues from the reference. It recognizes they're dinosaur-like and makes them move accordingly, for instance.
It understands, to a degree, that they're meant to be machines, and wheere it does add details, it's greebles rather than muscle or scales.
On the other hand, cars sometimes drive backwards. It's a rich tapestry, is what I'm saying.
The image(s) above in this post have not been modified/iterated extensively. As such, they do not meet the minimum expression threshold, and are in the public domain.
The image(s) above in this post have not been modified/iterated extensively. As such, they do not meet the minimum expression threshold, and are in the public domain.
Using vidu's reference feature with pixel art produces, at least in these instances, with an interesting pseudo-3d, pseudo-pixelart effect. While I was at first disappointed that it hadn't reinterpreted the concept into a more realized design, I've come to like the effect, and will probably use it in the video for SiniStar Meltdown.
The image(s) above in this post have not been modified/iterated extensively. As such, they do not meet the minimum expression threshold, and are in the public domain.