Ha, looks like we ended up tackling the same issue (and the same Reve vs. GPT-4o take) from slightly different perspectives today! We even reach the same broad conclusions. Mine's here: https://www.whytryai.com/p/openai-4o-native-image-generation
I've been playing around with 4o image generation and am honestly blown away. The level of detail it can pick up in an uploaded image and then re-render in any style is incredible.
Also, thanks for the shoutout to my Gemini 2.0 take!
I've been using Imagen 3 in Google Labs quite a bit - it's free and very solid. But now that GPT-4o is this incredible and I'm already a ChatGPT Plus user, this will likely become my go-to.
As I argue in my post, I just don't see too many people navigating to third-party text-to-image tools powered by diffusion models when ChatGPT is already a familiar interface. This probably goes for me, too, even though I'm more well-versed in image models than the average person.
The one thing that I still find really compelling about Midjourney is the web search feature. I don't consider myself a particularly creative person, and it's really fun to just type in different search terms and see a broad variety of what other people have dreamt up along those lines. But yeah, it's increasingly hard to justify paying for both.
You can search there, too, and there is just so much stuff people have been generating in Sora using 4o image generation. Enjoy!
Agreed - I'm already only paying for Midjourney sporadically to write the monthly prompts roundup, but I think tomorrow's might just be my last one. I've been running some of the Sora prompts in Midjourney, and the difference in prompt adherence is painfully obvious, especially for complex, lengthy prompts.
Ha, looks like we ended up tackling the same issue (and the same Reve vs. GPT-4o take) from slightly different perspectives today! We even reach the same broad conclusions. Mine's here: https://www.whytryai.com/p/openai-4o-native-image-generation
I've been playing around with 4o image generation and am honestly blown away. The level of detail it can pick up in an uploaded image and then re-render in any style is incredible.
Also, thanks for the shoutout to my Gemini 2.0 take!
With so many incredible model releases, what do you find yourself using on a daily basis?
I've been using Imagen 3 in Google Labs quite a bit - it's free and very solid. But now that GPT-4o is this incredible and I'm already a ChatGPT Plus user, this will likely become my go-to.
As I argue in my post, I just don't see too many people navigating to third-party text-to-image tools powered by diffusion models when ChatGPT is already a familiar interface. This probably goes for me, too, even though I'm more well-versed in image models than the average person.
The one thing that I still find really compelling about Midjourney is the web search feature. I don't consider myself a particularly creative person, and it's really fun to just type in different search terms and see a broad variety of what other people have dreamt up along those lines. But yeah, it's increasingly hard to justify paying for both.
Boy do I have news for you: https://sora.com/explore/images
You can search there, too, and there is just so much stuff people have been generating in Sora using 4o image generation. Enjoy!
Agreed - I'm already only paying for Midjourney sporadically to write the monthly prompts roundup, but I think tomorrow's might just be my last one. I've been running some of the Sora prompts in Midjourney, and the difference in prompt adherence is painfully obvious, especially for complex, lengthy prompts.