Post your AI images
Discussion
dinkel said:
Mr Whippy said:
I’ve been using getimg.ai and used pretty much 2,500 generations for 25 images for a website, (...).
There's no free sample and I need to use my Google account to get in > do you receive more spam?Yes pita using Google. I have an old account I don’t use but wasn’t keen at all having to use it to sign up (that said people use Google phones for their entire online lives soooo)
Leonardo is quite good too but has a convoluted setup via fecking discord.
It’s like midjourney using the web UI thing. Utterly daft.
Adobe will blow them all into the water in time if their best approach is having such daft accounting setups and interface logic.
I swear half the time it’s a way to make them feel a bit ‘alpha’ and the pita setup of an account or use largely makes you feel like an early entrant into test software and a bit special… but that window has already closed I think now Google and Bing etc have joined the party.
Luma AI can do the 'make a video from a photo' thing now. These show video versions of meme pictures like distracted boyfriend and Doge. https://x.com/MKBHD/status/1802371176141848764?s=1...
They’re still juddery, uncanny looking results, and the AI deployed is incapable of understanding why, despite being trained on likely billions of examples at this point.
Will feeding them billions more examples fix the fingers issues etc? Likely not.
And you can’t direct this without extreme pain. Ie, I like the face of the man, but not the woman, ok back to the still image, adjust using AI for 10 minutes, ok found one I like, now re-roll the video generation.
Oh, the entire video now does something different, I don’t like this now.
Any way to prompt the video to be the same but just change the face?
Nope, not yet.
How about have AI recreate the face on each frame of the previous data set?
But somehow have them go from a smile, to not smiling by the end?
Ok we’re now at the time of just filming this with real talent![hehe](/inc/images/hehe.gif)
Oh and we burned about £20 of energy away as heat and have 99.9% inefficiency as the bulk of the generated data is due to be deleted.
I do get all this stuff I really do, I’m now using it with clients but it’s all ideation side and rapid prototyping, or needs a lot of clean up.
When these AI can work with explicit prompting, great.
But so far the tags/prompts are clearly very loose and so the images out are as much use as a choccy fireguard if you have a direction in mind.
These AI need feeding with explicit tags with the training data.
Will feeding them billions more examples fix the fingers issues etc? Likely not.
And you can’t direct this without extreme pain. Ie, I like the face of the man, but not the woman, ok back to the still image, adjust using AI for 10 minutes, ok found one I like, now re-roll the video generation.
Oh, the entire video now does something different, I don’t like this now.
Any way to prompt the video to be the same but just change the face?
Nope, not yet.
How about have AI recreate the face on each frame of the previous data set?
But somehow have them go from a smile, to not smiling by the end?
Ok we’re now at the time of just filming this with real talent
![hehe](/inc/images/hehe.gif)
Oh and we burned about £20 of energy away as heat and have 99.9% inefficiency as the bulk of the generated data is due to be deleted.
I do get all this stuff I really do, I’m now using it with clients but it’s all ideation side and rapid prototyping, or needs a lot of clean up.
When these AI can work with explicit prompting, great.
But so far the tags/prompts are clearly very loose and so the images out are as much use as a choccy fireguard if you have a direction in mind.
These AI need feeding with explicit tags with the training data.
Gassing Station | Computers, Gadgets & Stuff | Top of Page | What's New | My Stuff