At any time when a brand new model of an AI software comes out, I attempt to take a look at it towards others and see how responsive (or helpful) they might be. However sure directions can nonetheless elude frequent sense with Generative AI – typically while you give it even easy instructions and instructions, it has a thoughts of it’s personal like an over-confident toddler headed towards a busy avenue.
I’m typically discovering myself arguing with the AI chatbots like OpenAI’s ChatGPT and Google Gemini Nano-Banana, as they may fairly reularly not solely botch-up the directions however they clearly don’t perceive them regardless of how clear and direct, simplistic and correct your immediate instructions are. Generally it’ll reply that it’s adopted the directions within the textual content repsonses however in actuality it has carried out NOTHING to the picture! So it brings little pleasure in arguing with it and enjoying good after a number of irritating fails… I often simply transfer on to a different software that can present me what I actually need.
I imply, look how far the instruments have quickly advanced since I first began reporting on AI imaging and Video Instruments again in 2023! (Keep in mind the HANDS?!!!) 😛
We’ve come to date in simply a few years, as predicted we’d, however now individuals are taking Generative AI instruments extra severely and producing much more content material than ever earlier than. Having these instruments at your disposal helps you be inventive and develop new concepts in little or no time. However to depend on them to do a whole video manufacturing is simply going to disappoint you, and your viewers.
Now that AI has been flooding our social and advertising and marketing channels over the previous couple years, typically individuals are on the lookout for cues that divulge to the very fact they’re really watching an AI video, so even the general public is getting extra eager to the phenomenon of the Uncanny Alley.
But many people have posted wonderful productions on-line that make it appear like AI is doing all this superior imaging, however in actuality, it’s no totally different than every other type of content material creation. For each one good consequence, there are probably dozens of failures behind that no one sees. A lot of the ensuing content material has been edited and filtered and massaged till it appears prepared for prime time.
What may go unsuitable?
The sudden, after all!
Right here, a easy product placement take a look at utilizing HeyGen’s Avatar IV with product placement. It appears certainly one of their default fashions has too many limbs and it’s utilizing ALL OF THEM! 😛
For this scene, I attempted each software simply to get the digital camera to pan/arc across the middle of this group – however ended up with additional folks and a lot spinning!
Lastly, some residence way of life makes an attempt so as to add folks shifting about to static residence photos… and take a look at the sneakers on the man on the patio with the hearth. I DID NOT ASK FOR THOSE! 😛
Scene 2, Take 247…
A pair initiatives that I’ve shared right here in my earlier opinions (all disclaimers nonetheless stand on these movies as define in every article).
From my article AI Instruments: Video & Animation Come to Midjourney, I needed to attempt a number of makes an attempt at every step to get the characters to reply to my prompts. Right here’s the ultimate rendered video:
Regardless of how the ensuing video(s) seem, there have been a number of FAILS alongside the best way at every stage. Listed here are only a couple immediate fails that didn’t make the lower.
From my current article AI Instruments: HeyGen Avatar IV Will get Actual, I attempted each AI Device I may get my palms on to create the varied motions and actions this AI mannequin was to make based mostly on the AI-generated script.
Right here’s the ultimate edited video (WARNING: PG-13):
Once I tried to instruct her to only “pull the set off on the drill” to make it spin, I acquired warnings in some AI instruments that the language was prohibited so having to creatively reword the prompts to attempt to get a consequence I wanted. You possibly can see that it simply wasn’t going to occur…
A number of different instruments and makes an attempt had been made and I lastly acquired sufficient to work with ultimately, however a few of these are simply too hilarious – proper as much as the tip! 😛
However simply keep in mind this…
THIS IS THE WORST THE TECHNOLOGY WILL EVER BE!
