Viewing sample resized to 177% of original (view original) Loading...
Description

Judy needs new pants every week!

Temporary Alts Link because Catbox is having issues: https://limewire.com/d/wpW5F#IVzoMxMwuQ

Alts: Alt1 Alt2 Alt3 Alt4

These were made with Grok Imagine for free and other people have been getting even better results. I'd encourage people to use it while it's free. Here's how I did this:

1. Sign up for an account at https://grok.com/ , you can use a Temporary Email like https://temporarymail.com/en/ but you might have to try a few different domains because some are blocked
2. Click on "Imagine" in the menu on the left hand side
3. Select "Video" in the input bar on the bottom
4. For text to video you can type the prompt now. For image to video you can upload the image, it'll automatically generate one video without a prompt, and then you can add a prompt from there.

Grok uses natural language prompting rather than tags. Unfortunately I lost the prompt for these videos, but what I had done was ask Grok text gen to make a template for optimal Grok Imagine gens and this is what it gave me:

[Main subject] [performs specific action/motion] in [environment if changing].
[Camera shot and movement, e.g. slow push-in, subtle handheld sway, gentle orbit].
Photorealistic cinematic style, exact face and body fidelity from the source image, preserve exact facial features, expression, lighting, and composition—no morphing or distortions.
Realistic physics, natural breathing, subtle eye movements, continuous 1x speed motion.
[Dialogue if any]: “Exact spoken words here.”
Audio: [sounds, e.g. soft wind, distant waves, gentle breathing, no music].

The amount of gens Grok lets you do for free seems to vary, but I was able to do many different attempts at this one tweaking the prompt a little bit each time. There's various other advice for prompts online if you search for it too.

Grok used to be uncensored but has since gotten more and more censored like other video gens sites, but some people have occasionally been able to get around the censorship. Here I did not use any explicit words in the prompt.

If you're doing an image to video gen having a good base image seems to make a big difference. The AI can pick up anything from the base image and go crazy with it. Here because the base image had some musk clouds / steam the AI really liked to over-emphasize it.

Hope that more people will experiment with it while it's free!

Blacklisted
  • Comments
  • There are no visible comments.