00:00I'm going to show you how I managed to make this anime using artificial intelligence.
00:05This is a big project that I am carrying out alone to be able to release my first episode.
00:10And as you can see, the characters are consistent.
00:15There is action, there is movement, there is even sound, music and of course voices.
00:20To achieve this type of result, I have spent about 40 hours of work so far.
00:25I'm going to show you step by step how you can do this.
00:29First, it is super important to tell you that you will have to define the story and of course define the graphic style of your anime.
00:38And to be able to start working on it quickly, I can recommend two AIs which are mid-day or of course AI design.
00:47And as you can see, I generated thousands of different images before I was able to achieve the style I wanted and have something consistent.
00:57There, if we look a little at midday, I find myself with a lot of drawings and generations in all directions.
01:06One of the big problems is of course always having characters who are going to be consistent.
01:11Luckily for that, on design, you will be able to train a character and summon it each time in your creations.
01:17And that will save you a lot of time.
01:19Once your character is trained, you simply select it and tell it to perform an action.
01:25To be able to create a character and train it, you will first have to generate what we call shit characters.
01:31That is, different positions of your character, behind, side, left, right, in action.
01:38You will generate these kinds of elements and you will take the images that are the most interesting each time.
01:44Typically, I like to work on a green screen because it allows me to cut them out more easily.
01:49and the AI understands much better that it is a character without taking into account the environment around it.
01:55Once you have the images you're interested in, you cut them all out one by one and give them to the AI to train.
02:02Then, you will be able to invoke this character in your creations.
02:06For example here, I asked him to jump in the air.
02:09And you can see every time I have pretty much the same type of character.
02:14The big problem is that it's not going to be perfect.
02:16And on top of that, you're going to have a definition that isn't very qualitative.
02:21So what we're going to do is we're simply going to take these images and give them to ChatGPT to correct.
02:27And that will be a point that is very important for you.
02:29As you see here, I will give ChatGPT the image generated by Design and of course also different positions of the character.
02:37Then I'll ask ChatGPT to correct the first image I gave it, while making sure to make the character look like the other images here.
02:47And as you can see, I iterated a lot of times.
02:50Because at first, this scene didn't suit me.
02:53Then I told him, it has to be dark.
02:56There needs to be a slightly blue, metallic environment like here.
03:00And as you can see, ChatGPT will then adapt and correct my image.
03:04Here, for example, I ask him to change the blue lights to a purple light.
03:09It is therefore a very long process that you will have to put in place to have quality images.
03:15And yet, it's essential because all the image generators you'll find each have their flaws.
03:21Personally, I have each image corrected by ChatGPT.
03:24Then I'm going to take those images and give them to Vidu.
03:28Vidu will be able to animate these different images according to an action.
03:32So here, for example, we have a zoom directly on the tank.
03:36Where here, I am having my different characters discuss.
03:39And again, as you can see, I have thousands of videos that have been generated.
03:44And from time to time, I generate them on a green screen so that I can replace the background with another background.
03:49But it doesn't stop there.
03:50I will also transform the different images into videos by midday.
03:54And as you can see, I have generated thousands of videos.
04:00And maybe sometimes even hundreds for a particular scene.
04:04Quite simply because it allows me to have the best every time.
04:08And that's a really important thing.
04:10Bulk generate so you can sort and select what works best.
04:15Don't expect to get the best result straight away.
04:18Then to generate the sounds I went to Eleven Labs and also Adobe Firefly with the sound generation feature.
04:26Then I told ChatGPT what sounds I wanted and it turned them into prompt text, which I then gave to Eleven Labs.
04:33I always work this way because ChatGPT or Gemini will optimize my prompts to get the best results.
04:41So, in reality, I save time and improve quality.
04:43As for the music, I went to Suno and generated thousands of different songs there too until I found the right one.
04:52And finally, to finish, to be able to adapt my voice, I simply recorded all the texts with my microphone.
04:59And I used what they call a voice changer at Eleven Labs.
05:02This way I can make my female characters, male characters, whatever, speak.
05:07And then, of course, I took all the different rushes from my videos and put them together over the music, over the voices,
05:16as you can see right here, like a classic video montage.
05:20And of course, during the editing, I had to repeat maybe 4 or 5 times to be able to refine the transitions,
05:37have similar color styles.
05:40And there, we are at approximately 70% of the final result.
05:44Once I finish the first episode, I'll re-export everything so I can convert it to 4K format with AI.
05:53If you want to know more about really all the prompts, about how I achieved all this,
05:58simply write “IA” in the comments.
06:00And there, through a text document, you will really have everything to follow step by step to be able to create your own animated film.
06:07And if you have made something using this tutorial, please comment it, I will be super happy to see it.
06:14In any case, I wish you a good creation and take care of yourself.
06:18Hello!
Commentaires