Although this is not an album, but merely a cover song, I can say without a doubt that I could not have done artwork this good if I was given weeks to make it. For informational purposes, I am going to display what my prompts were to show how I arrived at the end result. LMK if you have any questions.
I first did a query using the title of the song I was covering (Starfish & Coffee by Prince - my favorite song from him and NOT a hit). I did not use Prince in the command, just starfish and coffee. Here is the 4 pic grid I was provided with:
Pretty cool stuff and could have stopped there and worked with it, but decided to take some of the very visual lyrics of the song and search with that. The query prompt was "starfish and coffee, maple syrup and jam, butterscotch and a tangerine and a side order of ham" and I got this:
I liked the image in the lower right the best, so I then said give me 4 variations on that image and it produced this:
Of those, I liked the one in the upper left the best, so I said give me a higher resolution/quality version of that one and I got this:
I then took that image and put it into a template that I use for my "Studio Live" cover songs. I put some text over the top for date, copyright, title and original artist and this is the final result (image links to the page on my server where this stuff is fed from if you would like to listen - this song is first on the page):
Generating the images took a few minutes. Choosing my fonts and putting that beautiful artwork into my template took me 15 minutes or so, but I believe it resulted in a piece that otherwise would have taken me MANY hours even if I could have conceived such beauty.
Here are some things I have learned.
1. It does NOT do text well. Many images with have text that is not readable, and what is readable often does not make sense. There is a huge discussion on the Midjourney system that talks about this and ways to get your images to show no text and/or get accurate text on your images. There is not, at this time, a definitive method to get exactly what you want when it comes to the text on your images.
2. You can get a free account, and with that account they say that you can produce about 25 images. I produced 51 with my first free account. They also say in the getting started material, that they know some people will just set up another account with a different email address, but they encourage you to eventually purchase a subscription. I am on my 3rd free account but I will certainly be purchasing a subscription as the results I have already received would constitute countless hours of work on my part to produce such material.
3. When I do purchase an account, I will be going with the pro version. The major reason is, from best I can tell, that is the only subscription plan that allows you to render your stuff in privacy. All other plans (free, basic and mid) have you working in a "chat room" where others in the same room see your work. Most of what I will be doing is artwork for songs that are not yet released and as a result, I would like privacy in developing that rather than exposing my thoughts and artistic ideas to the public. Another thing is, all paid subscriptions are granted general commercial license to use the produced images however they see fit. Not true with the free account from what I can tell. They ask that any corporate entity with >$1,000,000 in annual revenue purchase the Pro version.
4. The different plans are based on the amount of "render" time you use. This is why they estimate that you will get about 25 images from the free account, which allows you 0.4 hours of render time before account gets locked. The paid plans increase the amount of render time and the features available. So with my first free account, I got 51 images, second account 46 images, third account is still running. this is because the images I generated did not use as much rendering time as others might and because I did not use any advanced quality render settings.
5. Just sitting in one of those chat rooms though can be inspiring both in what is produced as well as the commands used to produce it. There are a lot of variables/parameters that can be used to get specific looks/feels to the produced images, and I have learned a lot by looking at what others are producing and what they used in their parameters to get such results. That said, the simplest commands can produce amazing stuff, but when you get into it, you start to realize that you might want to go in other artistic directions and these parameters help you to do so.
6. You may have noticed in the commands above and the images produced, that I was not served a side order of ham in my requests. That is one way that the system falls short on occasion. Only one of the images produce maple syrup as well, but I ain't complainin', just sayin', there are limitations on the accuracy in addition to the text problem.
As mentioned above, if you have any questions, I will be happy to answer as best I can. Many have talked about the loss of work for artists, and although that is true to some extent, there is still the fine tuning and display elements that need to be added. It is also a great way for artists to get inspirations and look at things different at times, so it is also an assistant to artists. Peter Gabriel recently did an interview where he talks about AI and said "we are not going to stop it, so we may as well dance with it"! I agree!
Happy AI'ing all!
