chatgpt is really popular, and chatgpt has produced an epoch-making significance-since chatgpt, AI is really going to land.
There are so many things that chatgpt can do, even the programmers who originally developed the model themselves can't explain clearly what chatgpt can do. It seems that as long as you can think of it, it may help you.
Today I will tell you how to use chatgpt+midjourney to generate a fantasy animation video in 5 minutes, so that an ordinary person who has no writing experience and no drawing skills can also make his own animation story. The method is very simple, but the skills and experience are what I want to tell you.
Well, without further ado, let's start our tutorial.
In the first step, use midjourney to draw the protagonist's appearance, we use kunkun's photo as a benchmark.
Here is a brief introduction to midjourney, which allows you to use some scene descriptions and prompt words to generate corresponding pictures, such as:
(Key words: Biden and Trump play badminton together)
(Keywords: SpongeBob joins the Avengers)
You can even specify the picture style, author genre, canvas material, etc., which is very powerful.
Open the dialog box with the midjourney robot, click the plus sign in the lower left corner, and upload the real photo of the protagonist.
Use the /imagine command in the dialogue window with the midjourney robot, drag the uploaded picture to the prompt to copy the link, and enter the style and parameters you want to generate. For example, I fill in realistic, cartoon --ar 2:3 here, which means a realistic cartoon with an aspect ratio of 2:3. Now wait for it to be generated.
Among the generated pictures, I think the last one is the best, so we can click v4 below, which means to generate a group of similar pictures based on the fourth picture, which can help us choose better pictures. After generation, we select u3 again, which means we have selected the third image to generate a large image.
In the second step, we will use chatgpt to write a story
We can give a story outline and let chatgpt help expand it. I show an example here. There are mainly two characters, the male protagonist Kun and the female protagonist Shiba Yoshimiko, and then write the plot and dramatic conflict.
Now, open the chatgpt page, we start to hypnotize chatgpt, tell it that it is a novelist now, let it help us continue to write the story.
We translated the written outline into English, and after confirming that there were no translation problems, we posted it to the chatgpt dialog box and waited for it to be generated.
Now that we have a detailed story outline, let's see what the story is about. We can find that chatgpt expands our outline more completely and very logically.
The third step is to divide the story written by chatgpt into multiple scenes
Only in this way can midjourney be provided with a prompt to generate pictures. We tell chatgpt to break the above story into 20 scenes. Afterwards, we tell chatgpt to generate prompts for each scene, and if it misunderstands anything, we can always correct it.
Here is a trick for you to give chatgpt some examples. For example, I asked it to add fantasy style, realistic --ar 16:9 after each prompt, but he actually used brackets to enclose it. Then we can give some specific examples so that he can correctly generate the results we want.
The fourth step, use the generated prompt to generate an image in midjourney
We paste the prompt into midjourney. After the image is generated, we can select a certain image as the basis to continue generating in the previous way, or obtain a refined large image of a certain image, or even choose to completely regenerate it. Finally, we selected a picture, but there is a problem here, what should I do if the characters in the picture don’t look like Kunkun?
midjourney provides a fusion function, put the link of the prototype picture and the link of the character in the prompt, and then copy the prompt just now, so that the two pictures will be merged. Since the name Kun appears in our prompt, we need to indicate that this is Kun in front of the character portrait link. The generated image was not very satisfactory, we chose to regenerate a version, and finally, we chose the second image as the result.
The fifth step, we need to add dubbing to the animation
Here is an online website called Voicemaker, in which the synthesized human voice is very simulated. We choose Neural TTS, which makes the voice more natural. Then we choose a desired human voice, paste our story, and then we can generate and download mp3.
Well, here we have all the scene animations and sounds, open the editing software, import all the resources, add some effects and subtitles to complete our exciting animation, I named it "Kun's Fantasy Adventure Journey"
In fact, no matter chatgpt or midjourney, how powerful they are depends entirely on how we use them. Just like the Dragon Tuning Knife, beginners who don't know the secrets of martial arts can easily cut vegetables with the knife, but the masters can exert greater power of the Dragon Tuning Knife. For example, beginners using midjourney may input some simple words to generate some pictures, but experienced developers can generate very artistic pictures by adding some specific parameters. midjourney also shows some popular pictures.
This is what we need to constantly explore and learn. As a black-box model, humans can no longer systematically analyze and summarize its functions, but this is what excites us. You may discover functions that others have not discovered every day.