Midjourney: For the Artists

First up, Mid­jour­ney. If you've seen those incred­i­bly detailed, artis­tic, and some­times weird­ly beau­ti­ful AI images online, there's a good chance they were made with Mid­jour­ney. This tool is known for pro­duc­ing high-qual­i­­ty, aes­thet­i­cal­ly pleas­ing images. It has a spe­cif­ic "look" that many peo­ple find appeal­ing.

Who is it for?
Mid­jour­ney is for peo­ple who care about the final look of the image. It's great for artists, design­ers, and any­one who wants to cre­ate some­thing that looks like it was inten­tion­al­ly craft­ed. If you want to make con­cept art, illus­tra­tions, or just real­ly nice-look­ing pic­tures, this is a strong choice.

How it works:
This is the part that trips some peo­ple up. Mid­jour­ney oper­ates entire­ly through Dis­cord, a chat app. You don't go to a web­site to use it; you join their Dis­cord serv­er and type com­mands to a bot.

Here are the basic steps to get start­ed:
1. Get a Dis­cord Account: If you don't have one, you'll need to sign up for Dis­cord. It's free.
2. Join the Mid­jour­ney Serv­er: Go to the Mid­jour­ney web­site and you'll find an invi­ta­tion to their Dis­cord serv­er.
3. Find a "New­bie" Chan­nel: Once you're in the serv­er, look for chan­nels named some­thing like "#new­bies". This is where you can start mak­ing images.
4. Use the /imagine com­mand: To cre­ate an image, you type /imagine fol­lowed by a descrip­tion of what you want to see. This descrip­tion is called a "prompt". For exam­ple: /imagine a hyperrealistic image of a dog in a truck.
5. Wait for your images: Mid­jour­ney will then gen­er­ate four dif­fer­ent image options based on your prompt.

From there, you'll see but­tons labeled U1, U2, V1, V2, and so on. "U" stands for upscale, which cre­ates a larg­er, more detailed ver­sion of one of the four images. "V" stands for vari­a­tion, which gives you four new images that are sim­i­lar in style and com­po­si­tion to the one you select­ed. There's also a re-roll but­ton if you don't like any of the first four options.

What does it cost?
Mid­jour­ney is not free. You need to sub­scribe to one of their plans to use it. They have a few dif­fer­ent tiers, start­ing from a Basic Plan that gives you a lim­it­ed amount of image gen­er­a­tion time each month. The plans go up from there, offer­ing more "fast" gen­er­a­tion hours and the abil­i­ty to work in a "relax" mode, which doesn't use up your fast hours but takes longer to gen­er­ate images. The Basic plan is around $10 a month, with Stan­dard and Pro plans cost­ing more. If you pay for a year upfront, you get a dis­count. One thing to know is that on the cheap­er plans, your images are pub­lic by default. If you need pri­va­cy, you'll have to pay for the more expen­sive Pro plan, which includes a "Stealth Mode".

My expe­ri­ence with Mid­jour­ney is that it's excel­lent at inter­pret­ing style. You can tell it to make some­thing "in the style of Van Gogh" or "like a still from an ani­me," and it does a fan­tas­tic job. It takes some prac­tice to write good prompts, but the com­mu­ni­ty on Dis­cord is a good place to learn, as you can see what prompts oth­er peo­ple are using to get their results.

Stable Diffusion: For the Tinkerers and Tech-Savvy

Next is Sta­ble Dif­fu­sion. This one is dif­fer­ent from the oth­ers because it's open-source. That means the code is pub­licly avail­able, and any­one can use it, mod­i­fy it, and build on it. This makes it the most flex­i­ble and cus­tomiz­able tool on this list.

Who is it for?
Sta­ble Dif­fu­sion is for peo­ple who want con­trol. It’s for devel­op­ers, tech-savvy artists, and any­one who likes to exper­i­ment and get under the hood. If you want to run the AI on your own com­put­er, train it on your own images, and have fine-grained con­trol over every aspect of the image gen­er­a­tion process, this is the tool for you.

How it works:
Because it's open-source, there are many ways to use Sta­ble Dif­fu­sion.
* Run it local­ly: You can install it on your own PC. This requires a decent graph­ics card (GPU), espe­cial­ly one from Nvidia, as most mod­els are opti­mized for them. The set­up process can be a bit tech­ni­cal, involv­ing things like installing Python and using GitHub. But once it's run­ning, you have com­plete free­dom and don't have to pay per image.
* Use a web inter­face: Many web­sites offer a user-friend­­ly inter­face for Sta­ble Dif­fu­sion. Some are free with lim­i­ta­tions, while oth­ers use a cred­it sys­tem. Dream­Stu­dio is the offi­cial web inter­face from Sta­bil­i­ty AI, the com­pa­ny behind the mod­el.
* Cus­tom Mod­els: One of the biggest strengths of Sta­ble Dif­fu­sion is the huge com­mu­ni­ty cre­at­ing and shar­ing cus­tom mod­els. These are ver­sions of the main mod­el that have been trained for spe­cif­ic styles, like cre­at­ing ani­me char­ac­ters, pho­to­re­al­is­tic por­traits, or fan­ta­sy land­scapes. Web­sites like Civ­i­tai are mas­sive libraries for these mod­els.

The process of cre­at­ing an image is sim­i­lar to oth­er tools: you write a prompt. But with Sta­ble Dif­fu­sion, you also have a lot more para­me­ters to play with, like choos­ing the sam­pling method, adjust­ing the guid­ance scale (how close­ly the AI sticks to your prompt), and using neg­a­tive prompts to spec­i­fy what you don't want to see. You can also do things like "inpaint­ing" (regen­er­at­ing a spe­cif­ic part of an image) and "out­paint­ing" (extend­ing an image beyond its orig­i­nal bor­ders).

What does it cost?
The mod­el itself is free. If you run it on your own com­put­er, the only cost is the elec­tric­i­ty. If you use a web ser­vice, the pric­ing varies. Many offer a cer­tain num­ber of free cred­its to start, with options to buy more.

My take on Sta­ble Dif­fu­sion is that it has the steep­est learn­ing curve, but it's also the most pow­er­ful if you're will­ing to put in the time. The abil­i­ty to switch between thou­sands of com­­mu­ni­­ty-made mod­els is a huge advan­tage. You can find a mod­el for almost any style you can imag­ine. The trade-off is that it can some­times feel less pol­ished and more tech­ni­cal than Mid­jour­ney.

DALL‑E 3: For Simplicity and Integration

DALL‑E 3 is the image gen­er­a­tion mod­el from Ope­nAI, the same com­pa­ny that makes Chat­G­PT. Its biggest strength is its inte­gra­tion with Chat­G­PT and its abil­i­ty to under­stand nat­ur­al lan­guage incred­i­bly well.

Who is it for?
DALL‑E 3 is great for begin­ners and peo­ple who are already using Chat­G­PT. It's also for any­one who val­ues prompt accu­ra­cy. It's often bet­ter than the oth­ers at under­stand­ing com­plex, detailed prompts and gen­er­at­ing images that include spe­cif­ic ele­ments exact­ly as you described them. It's also sur­pris­ing­ly good at cre­at­ing images with leg­i­ble text, which is some­thing most oth­er AI art tools strug­gle with.

How it works:
The main way to use DALL‑E 3 is through Chat­G­PT.
1. You need a Chat­G­PT Plus, Team, or Enter­prise sub­scrip­tion. DALL‑E 3 is not avail­able on the free ver­sion of Chat­G­PT.
2. Start a con­ver­sa­tion: You sim­ply describe the image you want in the chat. You don't need spe­cial com­mands. You can just talk to it like you would a per­son. For exam­ple, "Cre­ate an image of a pan­da rid­ing a bicy­cle on the moon."
3. Refine the image: This is where DALL‑E 3 shines. Because it's inte­grat­ed with Chat­G­PT, you can have a con­ver­sa­tion to tweak the image. You can say things like, "Okay, now make the pan­da wear a red hel­met," or "Can you change the style to be more like a car­toon?". Chat­G­PT will then gen­er­ate new images based on your feed­back.

You can also access DALL‑E 3 for free through Microsoft's Bing Image Cre­ator, though there may be some lim­i­ta­tions com­pared to the paid Chat­G­PT ver­sion.

What does it cost?
To use it with­in Chat­G­PT, you need a sub­scrip­tion, which is about $20 per month. Using it via Bing Image Cre­ator is free. OpenAI's terms allow for com­mer­cial use of the images you cre­ate.

I've found DALL‑E 3 to be the most straight­for­ward of the big three. The con­ver­sa­tion­al approach to refin­ing images feels nat­ur­al and is very effec­tive. If I have a very spe­cif­ic scene in my head with mul­ti­ple ele­ments that need to be in the right place, I often turn to DALL‑E 3 because it's the best at fol­low­ing com­plex instruc­tions. The image qual­i­ty can some­times feel a bit less "artis­tic" than Mid­jour­ney out of the box, but its abil­i­ty to under­stand lan­guage is sec­ond to none.

Other Noteworthy Free Tools

If you're not ready to pay for a sub­scrip­tion or install soft­ware, there are some decent free options to get your feet wet.
* Can­va: Known for graph­ic design, Can­va now has a free AI image gen­er­a­tor that is very user-friend­­ly. It's a good option for begin­ners.
* Adobe Fire­fly: Adobe has its own AI art gen­er­a­tor. It's designed to be com­mer­cial­ly safe because it was trained on Adobe's stock image library. It has some neat fea­tures and is inte­grat­ed into Adobe prod­ucts like Pho­to­shop.
* Craiy­on: This was for­mer­ly known as DALL‑E mini and was one of the first wide­ly avail­able AI art tools. It's free to use and sim­ple, though the qual­i­ty isn't on the same lev­el as the big three. It's a fun way to exper­i­ment with­out any com­mit­ment.

Ulti­mate­ly, the "best" tool depends on what you want to do. If you want beau­ti­ful, artis­tic images and don't mind using Dis­cord and pay­ing a sub­scrip­tion, go with Mid­jour­ney. If you want max­i­mum con­trol, cus­tomiza­tion, and don't mind a tech­ni­cal set­up, Sta­ble Dif­fu­sion is your best bet. And if you want ease of use, excel­lent prompt under­stand­ing, and a con­ver­sa­tion­al way to cre­ate, DALL‑E 3 is a fan­tas­tic choice.