Midjourney and Dall-E: our tips on how to aim like a pro!

Halfway through the trip:

– Be as precise as possible : Generally speaking, the biggest error lies in the lack of precision. For optimal results, have a clear idea of the desired result in mind. The risk of error will be lower in a close-up image: the larger the field of the requested image, the more precise the request must be. Otherwise, the AI will take liberties when generating background elements.

– Never stop at the first two versions of the generated visual. :

You have to be patient to get a satisfactory version. As version 6 of Midjourney finally achieves consistency (being able to modify the same image at will), it is now possible to generate a character and put it in different situations, contexts or attitudes! Have fun and don’t hesitate to iterate with the platform to modify certain details and achieve the perfect result.

– Know how to choose your words and punctuation:

Don’t use too many opposite terms in the same sentence (small/big, sad/happy, etc.), otherwise the AI won’t offer an average of the two. Prefers the use of English to obtain more qualitative results. Finally, separate all your ideas or blocks of similar ideas with a period. Also make sure you put them in a logical order and don’t list them out of order.

– Consider using immediate assistance tools :

a good message is made up of a dozen parameters (lighting, framing, position, style, format, environment, color, mood, etc.). In order not to forget anything, don’t hesitate to be smart! For this, there are several solutions: for example, you can ask ChatGPT to generate the entire message of an image you send it or to describe what it sees there. You can also explore the Midjourney site (and the Discord forums): the settings tab details the expressions and conventions to use to create good messages.

– Consider avoiding inclusion bias : With AI driven by predominantly Western content, the proposed results rarely take into account issues of inclusion. If you want an image that includes ethnic diversity, remember to specify it. The same goes for gender diversity.

– Dare to innovate to produce something new. : Don’t make the mistake of imitating what already exists. The possibilities of Midjourney are endless, use them to create unique and exclusive images by combining various sources of inspiration.

South Dall-E:

– Consider giving context : To give Dall-E a better idea of where you want to go, you can give him a little context: what kind of project do you need to create this visual for? What is the concept of the campaign or support to be carried out? What objectives will be achieved? The result will be even more relevant if you define a role: for example, “You are an artistic director at a Parisian advertising agency.” However, be careful to distinguish the context information (your needs for your project) from the message itself (what you want to appear in the image).

– Use 5W (What, Who, When, Where, Why): these 5 elements constitute a good basis for establishing the spatio-temporal framework of the desired visual.

– Do not request tickets in the visual. : Requiring the visual to contain writing increases the risk of error, since Dall-E generally does not spell words correctly. Also be careful with homonyms, which you do not always know how to interpret correctly.

– Asks to maintain graphic coherence. – In Dall-E, when you have generated a visual using a first message, sending a second message deletes the first generated visual. If you want to keep it to make changes, remember to specify it. For example: “Don’t change anything, but delete or add XXX.”

– Use Dall-E to propose visual intentions. : Dall-E allows you to generate creative intentions, rather than finished images. Feel free to send them visual inspiration to help them understand what graphic style you want to use.