Why are you dissatisfied with the drawings halfway?

The painting effect in the middle is always unsatisfactory, isn't it?

Is there a treasure like me? At first, I copied and pasted other people's keywords directly, and then I changed them, and there was a picture. The final effect is not satisfactory! Aside from the randomness of Mid journey, another reason is that we don't have enough control over our keywords. In fact, many keywords and grammars are of little use to the generation logic of Mid-Journey (invalid keywords and grammars), so we often generate some unexpected pictures.

In fact, the official grammatical structure has been given, so today I will explain the official grammatical structure of Mid journey in depth.

What is a hint?

"Hint" refers to the short text phrase used to generate pictures. Mid journey Bot will break down the words and phrases in prompt into smaller units called token, which can be compared with its training data and then used to generate a picture.

That is to say, no matter what sentence pattern, MJ will first disassemble long sentences, and then generate the final picture through the "black box algorithm" of multiple to kem, so the meaning of language order is not great.

MJ doesn't know grammar at all.

Mid journey doesn't understand grammar, sentence structure or words like humans do. In many cases, more specific synonyms are more effective. For example, you can use the words "huge" and "huge" instead of "big". Use as few words as possible. Using fewer words means that each word is more influential. Use commas and brackets to help organize ideas, but it should be noted that Mid journey will not explain them reliably. Bot is not sensitive to internal writing.

MJ official tips recommended

A simple hint may be just a sentence (or even an expression), and a picture can be generated.

A very short prompt will depend heavily on the default style of Mid journey, so a more descriptive prompt will produce a more unique effect. However, excessively lengthy hints are not always better.

Focus on the main concepts you want to express. Regardless of the specific or vague description, MJ will automatically generate pictures according to the description.

Vague descriptions may bring more changes, but you may not get the specific details you want.

Therefore, please describe any background or details that are important to you as clearly as possible. Consider the following factors: theme, style, environment, lighting, color, mood, composition.

MJ official tips recommend theme characters, animals, people, objects, etc.

Style photos, paintings, illustrations, sculptures, graffiti, tapestries, etc.

Indoor, outdoor, lunar and underwater environments.

Soft lighting, environment, cloudy days, neon lights, studio lights, etc.

Bright, soft, bright, monochrome, color, black and white, etc.

Calm, peaceful, noisy, energetic, etc.

Portrait, head, close-up, bird's eye view, etc.

MJ official tips recommended

Plus: plural words may bring a lot of randomness.

Please try to use specific figures. For example, "three cats" is more specific than "cats". You can also use collective noun, such as "a flock of birds" instead of "birds".

Finally, I'd like to share with you a half-way dictionary made by National Kelang. All the keywords were tested by the team, and thousands of commonly used words were carefully screened. 500+ practical keywords, including style, lighting effect, material performance and other elements.

The first version of the document is open source and free, pdf version. If you need to hide the treasure, please leave a message below. I will share it with you ~

Chinese-English dictionary of graphic keywords. Pdf is made by crown. All keywords are tested by the team, and 500+ practical keywords are carefully selected, including style, lighting effect, I material performance and other elements.