I need to remember to just come here one day when I'm generating an image (so, most days) and experiment w/one of these. The 16 bit one is a gold mine for the type of stuff I'm writing about the most this year, and the others are handy for anyone stuck with a meh image.
I like taking a step back sometimes and reaffirming to myself that text-to-image models aren't a homogenous group at all. It's fun to see how much variation there is for the same exact prompt.
I've played the game w/identical prompts for Gemini and ChatGPT (and Dall E "back in the day"), and even that little side by side is instructive as to how the models "think."
It was on my shortlist along with Recraft V3 and Stable Diffusion 3.5, but I wanted to keep things manageable. If you do end up trying these in Leonardo, I'd love to see the outputs!
I have been trying more with Midjourney recently and, thanks to you principally, it’s been quite a lot of fun. Thanks again, Daniel. All the best, John.
That's awesome to hear John! What kind of stuff do you create in Midjourney? If you ever feel like sharing, feel free to send over some of your favorite images!
MidJourney definitely has been one of the most amazing tools to get into AI. I wish there were ways to control it a little bit better though. How’s the text coherence these days? Since Ideogram came out, I haven’t been subscribed.
Yeah, I've been a Midjourney fanboy since at least the V4 days and have written dozens of posts about it. But text coherence was never its strong suit. You can check out my relatively recent test here: https://www.whytryai.com/p/ai-image-model-spelling-text
Things haven't gotten drastically better for Midjourney when it comes to text. But they've introduced lots of editing and restyling tools that might be worth checking out. The Midjourney team seems to be prioritizing things like personalized aesthetics (the platform learns what you like and customizes the look of its images based on your preferences).
That feels like the right move, seeing how GPT-4o makes it very hard to compete on things like scene coherence, text accuracy, etc. Midjourney might as well lean into its strength and position itself as an aesthetic exploration tool.
I need to remember to just come here one day when I'm generating an image (so, most days) and experiment w/one of these. The 16 bit one is a gold mine for the type of stuff I'm writing about the most this year, and the others are handy for anyone stuck with a meh image.
I like taking a step back sometimes and reaffirming to myself that text-to-image models aren't a homogenous group at all. It's fun to see how much variation there is for the same exact prompt.
I've played the game w/identical prompts for Gemini and ChatGPT (and Dall E "back in the day"), and even that little side by side is instructive as to how the models "think."
There's that word again!
Cool testing. Would’ve been curious to see outputs from Leonardo!
It was on my shortlist along with Recraft V3 and Stable Diffusion 3.5, but I wanted to keep things manageable. If you do end up trying these in Leonardo, I'd love to see the outputs!
Fascinating take on how to approach each Ai deck and includi g the music and other areas, always inspiring and a great quest, thank you
Thanks Daniel. I would need a lot more practice!
I have been trying more with Midjourney recently and, thanks to you principally, it’s been quite a lot of fun. Thanks again, Daniel. All the best, John.
That's awesome to hear John! What kind of stuff do you create in Midjourney? If you ever feel like sharing, feel free to send over some of your favorite images!
MidJourney definitely has been one of the most amazing tools to get into AI. I wish there were ways to control it a little bit better though. How’s the text coherence these days? Since Ideogram came out, I haven’t been subscribed.
Yeah, I've been a Midjourney fanboy since at least the V4 days and have written dozens of posts about it. But text coherence was never its strong suit. You can check out my relatively recent test here: https://www.whytryai.com/p/ai-image-model-spelling-text
Things haven't gotten drastically better for Midjourney when it comes to text. But they've introduced lots of editing and restyling tools that might be worth checking out. The Midjourney team seems to be prioritizing things like personalized aesthetics (the platform learns what you like and customizes the look of its images based on your preferences).
That feels like the right move, seeing how GPT-4o makes it very hard to compete on things like scene coherence, text accuracy, etc. Midjourney might as well lean into its strength and position itself as an aesthetic exploration tool.