AI Videos
02/12/26: Okay, I tried the updated Sora and it's pretty good. All I gave it was these images and the simple prompt "Create a video with these two people walking through a forest in New York state in Autumn." It made the dialog up on its own, lol. This is wild, though, when you think about it. I asked Gemini to create a Regan based on a text description. The Trixie is made from the cream oval Trixie. Now Sora took that and turned it into walking, talking people.
I discovered why the DeeVid AI turned Honey and Trixie into college-age looking girls in the horseback riding videos below. It will not create videos with children or teens. Interestingly, the reasons it gave didn't apply to what I was asking, but it explains why no matter how hard I tried, it didn't work.
I discovered why the DeeVid AI turned Honey and Trixie into college-age looking girls in the horseback riding videos below. It will not create videos with children or teens. Interestingly, the reasons it gave didn't apply to what I was asking, but it explains why no matter how hard I tried, it didn't work.
|
|
|
|
|
Using the same headshots, I tried this prompt: Create a video of these two people standing in a horse stable. They're talking about whether or not the weather will be nice enough for a late afternoon ride through the game preserve. The woman is shorter than the man.
And yes, I had to add that she was shorter than him because in the first pass, she was a couple inches taller, lol! Random thoughts: It's interesting how fast the AI makes the characters move and talk. Also? Apparently it has deer on the brain because it mentioned them in both vids. |
|
|
Prompt: Create a video of these three people standing in front of a spooky, abandoned Victorian mansion. Dialog: Have the blond young man on the left say, "Are we sure we want to go in?" Dialog: The young woman says, "Now that we're here, we might as well look around." Dialog: The dark haired young man on the right says, "Sarah Sligo's ghost or not, I have a job to do, remember?"
Oh, so close, Gemini! Missed it by *that* much. XD
(The pic was made from images used in 31 Things to Do in Sleepyside When You're Dead.)
Oh, so close, Gemini! Missed it by *that* much. XD
(The pic was made from images used in 31 Things to Do in Sleepyside When You're Dead.)
01/25/26: It definitely still needs work, but it is absolutely wild how far this has come in the last two or three years. These videos were made using Gemini, DeeVid, Sora, and Flow. I think by the time I retire in 14 years, I may be entertaining myself by turning ASKOF into an actual television show. *snort*
Note: You can click on the "expand" arrows in the videos if you want to make them larger.
Note: You can click on the "expand" arrows in the videos if you want to make them larger.
|
|
Ok, this one is killing me. I uploaded the still shot of Trixie Gemini had created and used this as a prompt:
"Can you create a video of this young woman running down a dark forest path trying to get away from someone off screen who is chasing her?" The dialog it came up with on its own is hysterical. Also, pay close attention to what happens with her body and head the first time she turns around. "Help! Help, please! Don't bother. Leave me alone! Not a chance." |
|
This one is much better. Probably because I used a better prompt:
"Can you create a video of this woman in a spooky old graveyard at sunset? She approaches an old tombstone and slowly bends down to brush weeds away from it so she can read it. Can you just have eerie music playing but no dialog?" |
|
|
So Gemini created this picture for me for Trixie and Regan traipsing through the preserve. He's too short (or she's too tall) and his hair isn't red enough, but I went with it.
The prompt for this first video was: "Can you turn this into a video with the couple walking? She stops to turn and look at something off the path and he looks at her in concern. No dialog, but include a soft melancholy music." Kinda missed a bit, Gemini! |
|
I gave it a new prompt:
"Can you redo this video where she looks away from him, like she's seen or heard something off in the woods, and he looks at her in concern?" And we get unnecessarily creepy Regan for our second video. |
|
|
|
And then I tried, this prompt:
"Okay, can you redo this now to add dialog? Have her give a small gasp and say, 'Did you hear that?" Have him reply, "No. What did you hear?'" LOL, so close, Gemini! |
This next one is pretty good. I used a freebie trial at DeeVid for it. I started with the picture of Trixie from the home page here and gave it this prompt:
"Please make a video of this woman walking calmly through the woods with typical ambient sounds. Suddenly, she stops and gasps, looking back over her shoulder because she thought she heard something. There is a loud crack like someone stepped on a tree branch and a crow caws."
It didn't quite follow instructions, but it was close. The interesting things about DeeVid, if I am understanding it right, is you can stitch together a longer video. A prompt asked me if I wanted to "extend it."
"Please make a video of this woman walking calmly through the woods with typical ambient sounds. Suddenly, she stops and gasps, looking back over her shoulder because she thought she heard something. There is a loud crack like someone stepped on a tree branch and a crow caws."
It didn't quite follow instructions, but it was close. The interesting things about DeeVid, if I am understanding it right, is you can stitch together a longer video. A prompt asked me if I wanted to "extend it."
|
This one also from DeeVid got the dialog right, but Regan's hair is now just full on dark brown and also, he sounds more like he's exasperated with Trixie than concerned.
The prompt was: "Please make a video of this couple walking through the woods at sunset, with typical forest sounds. She stops and looks off into the distance, asking, 'Did you hear that?' He shakes his head and says, 'No, what did you hear?'" If I am learning anything from these experiments, it's that I really need to take a crash course in AI prompting, lol. |
|
|
I have no idea why the girls were changed!
The prompt was, "Please make a video of these girls riding along a forest trail. One says, 'We should probably head back soon.' The other says, 'Ok, but let's check on the clubhouse first.'" |
|
|
But wait! It get's better. I changed the prompt to say:
"Please make a video of these girls riding along a forest trail. One says, 'We should probably head back soon.' The other says, 'Ok, but let's check on the clubhouse first.' Please make sure the girls are the exact ones in this photo." Aaand, it did... this. Yeah, not what I said, DeeVid! |
|
|
|
I started with a screenshot of the yellow cover of The Whispering Witch, then asked it to create a video where the witch points at Trixie and slowly fades away, and I got this. Not sure why the witch acts like she suddenly got tased before disappearing, but I find the whole thing pretty funny.
|
|
It gets better, though, because when I asked it to redo the video but add spooky music to the background, it instead gave me this.
Like, seriously. What? No one threw a bucket of water on her, hahaha. |
|
|
|
This one is brought to us by Sora2, with the prompt, "Can you have this young woman walk into a stable and call out, 'Regan? Are you here?'"
The "REEgan" people should be pleased, lol. |
|
Sora2 did better with the whispering witch, though it's interesting that Trixie is frozen in place.
|
|
|
|
Gemini did better with Regan in creating the image used for this. And Sora2 did ok, though it switched the speakers from what my prompt was. Regan has Trixie's line and she has his.
|
|
I gave the graveyard another try, changing the dialog a little and starting with an altered image. I also had Gemini remove the red frame first.
And the role of Regan is now being played by a nice, clean-cut Englishman. *dies again* I especially like that he waves vaguely to the right while saying it's off to the left... |
|
|
|
Still with Sora2. I find it interesting that this is the second video where Trixie raises her hands to cup her mouth like she's about to call out, but then drops them before she says anything.
Also, now we have "Raygun." I didn't tell it either time how to pronounce it. Not sure why her hair looks different this time around. |
|
These next videos, I switched to testing Flow. At first, it kept glitching with English accents even when I specified American. This video is ok, but Trixie's cheerful smile at the end is out of place.
|
|
|
|
This one, using the same prompt, has a better tone, but apparently Trixie is a ghost, because she passes right through a gravemarker.
|
|
And now, an ASKOF scene. Flow made it waaaaay more dramatic than it was supposed to be.
And I love that the bats look like oversized rubber bats you might find at Spirit Halloween. |
|
|
|
And another ASKOF scene. At least this one isn't wildly over dramatic, lol.
So, what I learned with my experimenting today: Generally, AI does pretty good with taking the people in still images and creating videos from them. Even if you specify an American accent, you might get British. Although, in one of my attempts, Regan came out sounding like a southern cowboy, and if you listen closely to this last video, you can hear a hint of it. Overall? This is amazing. Just a few short years ago, AI was struggling with getting eyes and hair right and giving people seven fingers on one hand in still images, and now we're already making videos. I really do think one day the technology will be there that I could actually create an entire ASKOF television episode, which totally blows my mind. |