generated by AI I've reached the point in my life where I'm getting ideas for bits and skits, but I'm not really writing them out. It's not like I have an outlet. I mentioned elsewhere that I've been running some of my song parodies through an AI song creator. I get it, AI isn't a creative force, but it is a tool. But a stupid one. For instance, if I write lyrics to the tune of "Born to Run," there's no way to tell the AI thing that's what I need it to do. And, as a comic device, song parody is very much made funny by the fact that you're messing around with an actual song (check out Randy Rainbow). At least now I have something tangible to put on display.
Then I had an idea, it's a quick one. Practically a black out. It would make a decent commercial parody. But my old radio outlets are gone. I don't have a posse of cohorts that can get together and just film stuff on the fly. Then I got an idea...AI. Why don't I run my script through AI and see if I can "produce" the sketch myself. Easier said than done.
No, I can't just "upload" the script and it'll produce a video. Also, it costs. I don't have a lot of "disposable income" to pump into the internet. I see a lot of AI slop popping in my various timelines and walls, and most of it is 5-second nonsense. Some of it looks stupid and makes me wonder why someone would post it on purpose.
I looked some stuff up and, really, you have to piecemeal anything of length together. Which isn't a major issue. I've been using edit software for ages, to put together highlight reels, grab my produced work and add credits and details. AI is basically 5-second clips that I can generate, then I'd have to edit them together. Here's the rub: AI is really inconsistent. There's no finesse. There's no tweaking. There's no guarantee AI will even follow your written instructions. Say "remove confetti" and it will fill the image with confetti. Say a character grabs something and it will grab something with the hand holding a cell phone and somehow merge the object into a weird mutant thing. And it doesn't have a memory. Even if it comes out close to what you wanted, you just can't refine it. If you want to "regenerate" a video, it starts from scratch. Nothing looks the same, there's no continuity. It's a clean slate every single time and that's a tough thing to work around.
You have to have AI create some characters in the "long shot" of the location they'll be in. Then I have to edit and crop that image to extract my characters. Now, if I want them to talk, I have to run them through another program to make the image move and speak. BUT I have to create an audio track with some AI voices and I have to do all the dialogue for each character individually. Plus, the video creation will simple stop at the 15-second mark, forcing you to create multiple tracks for a single character. You run that and that's what you'll be making clips from, hoping the character remains the same from clip to clip.
I edited and used pan and scan to work around the gaffes (and watermarks).
Oh, you get a lot of gaffes. That's the trick with AI sites. They charge you for each bit of content you generate, even if it's awful or wrong. So you eat through credits like Donald at a Mickey D's. I've been jumping between multiple sites, using free trial periods, earned daily credits and one site I actually subscribed to for a minimal number of creations. And don't get me started about picture ratios. Focusing on one character, having cropped it from a larger photo, means each scene will be a different screen size. So that's distracting. And I don't have the resources or brainpower to work through it.
The product is a real mixed bag and it's not creative in and of itself. However, it gives me a new way to put my material out into the world.
I played with some material and AI earlier (Abbott & Costello Elementary, Flood Country) but this was the first attempt at doing a straight-through, linear video. The word turducken comes up a lot this time of year and suddenly I heard it differently this time and decided that donut company should be all over it.
Here it is, in time for the holidays: the new Tur-Dunkin' coffee flavors for Thanksgiving. I've gotten some good feedback (For the writing, not the AI. Then again, that's kind of the goal).
No, I can't just "upload" the script and it'll produce a video. Also, it costs. I don't have a lot of "disposable income" to pump into the internet. I see a lot of AI slop popping in my various timelines and walls, and most of it is 5-second nonsense. Some of it looks stupid and makes me wonder why someone would post it on purpose.
I looked some stuff up and, really, you have to piecemeal anything of length together. Which isn't a major issue. I've been using edit software for ages, to put together highlight reels, grab my produced work and add credits and details. AI is basically 5-second clips that I can generate, then I'd have to edit them together. Here's the rub: AI is really inconsistent. There's no finesse. There's no tweaking. There's no guarantee AI will even follow your written instructions. Say "remove confetti" and it will fill the image with confetti. Say a character grabs something and it will grab something with the hand holding a cell phone and somehow merge the object into a weird mutant thing. And it doesn't have a memory. Even if it comes out close to what you wanted, you just can't refine it. If you want to "regenerate" a video, it starts from scratch. Nothing looks the same, there's no continuity. It's a clean slate every single time and that's a tough thing to work around.
You have to have AI create some characters in the "long shot" of the location they'll be in. Then I have to edit and crop that image to extract my characters. Now, if I want them to talk, I have to run them through another program to make the image move and speak. BUT I have to create an audio track with some AI voices and I have to do all the dialogue for each character individually. Plus, the video creation will simple stop at the 15-second mark, forcing you to create multiple tracks for a single character. You run that and that's what you'll be making clips from, hoping the character remains the same from clip to clip.
I edited and used pan and scan to work around the gaffes (and watermarks).
Oh, you get a lot of gaffes. That's the trick with AI sites. They charge you for each bit of content you generate, even if it's awful or wrong. So you eat through credits like Donald at a Mickey D's. I've been jumping between multiple sites, using free trial periods, earned daily credits and one site I actually subscribed to for a minimal number of creations. And don't get me started about picture ratios. Focusing on one character, having cropped it from a larger photo, means each scene will be a different screen size. So that's distracting. And I don't have the resources or brainpower to work through it.
The product is a real mixed bag and it's not creative in and of itself. However, it gives me a new way to put my material out into the world.
I played with some material and AI earlier (Abbott & Costello Elementary, Flood Country) but this was the first attempt at doing a straight-through, linear video. The word turducken comes up a lot this time of year and suddenly I heard it differently this time and decided that donut company should be all over it.
Here it is, in time for the holidays: the new Tur-Dunkin' coffee flavors for Thanksgiving. I've gotten some good feedback (For the writing, not the AI. Then again, that's kind of the goal).
