it will be a new term
it will surpass the historical peaks of stable diffusion & midjourney visibile in the image below, on the same y axis
i.e. when it comes out, the total interest in it will be higher than the total interest was in those prior two systems at their peak
1,000
3.00it will surpass the historical peaks of stable diffusion & midjourney visibile in the image below, on the same y axis
@Ernie There is now a higher Midjourney peak in April 2023. Based on the title, I assume for this to resolve "YES" the benchmark to beat is the April 2023 peak, not the 2022 peak shown in the description?
Models currently capable of text are FLUX.1, Stable Diffusion 3.5, ReCraft v3, and Ideogram.
It doesn't look like any of them are going to unseat Midjourney at the peak of its hype:
@Ernie Are you looking to judge this based on Google Trends for the US only, or worldwide? The screenshot in the description restricts the search to the US.
@SergeyDavidoff tough questions. Bfl and ideogram destroy mj but the hype isn't as big. I'll look into yuur questions but I'm also hoping the answer becomes obvious
@Ernie I've checked both US and worldwide trends, and right now the outcome is the same in both regions. So the region question doesn't matter as much.
The critical question is whether we're comparing against the 2022 peak of Midjourney or the April 2023 peak. Based on the title I assume it's the latter, but I'd like to have a clear answer when so much mana is on the line.
I believe GPT-4o has native image generation capabilities & accurate text
@SaviorofPlant agree on the tech based on the demo images.. this claim is specific about how to evaluate it so I'm waiting. Hopefully it comes out to the API soon
@Ernie How would you measure the trends data for this? Most users are probably going to be searching "ChatGPT" and not "GPT-4o"
@SaviorofPlant yeah. The claim is fairly focused on hype level of the image generation tool, so it's not about absolute image quality actually, but rather being able to measure a high hype level.
Measuring it by searches for chatgpt would incorrectly include general gpt info, so that doesn't seem wise.
I'm open to ways to fairly measure this including combining search terms
