Today I'm sharing the followingAI Video CaseI don't know what you're talking about, but I think it's a good idea to talk about how many people are interested inshape of mouthHow to do it.
Tools used
I.e., dreams:https://www.1ai.net/10005.html
Can be spiritual:https://www.1ai.net/12558.html
Conch:https://www.1ai.net/21428.html
lip synching
This shot is of a girl with a cell phone in her hand beingSelfie Vlog, in order to reflect the effect of this selfie, to add keywords such as first view, selfie view, etc. inside the cue, the ratio of 9:16 or 2:3 vertical screen can be.
The prompt I came up with myself was: college girl walking down an old mountain trail, taking a selfie, short hair, white t-shirt, blue jeans, just after a light rain.
I then threw the cue word to the big model to optimize my cue word and output it as a professional detailed AI drawing cue word, so the big model gave me a professional cue word, which was well written and generated acceptable results.
Split Lifetime Chart Prompt Words
FIRST POV, FIRST PERSON SELFIE POV, FRONT SIDE, FACING THE CAMERA, a young female Chinese college student in her early 20's, short hair, wearing a white t-shirt, blue jeans, and an outdoor backpack, young woman, is holding up a selfie stick and smiling brightly at the camera. She is on an old, mist-shrouded stone road in the mountains of China. The light is a soft, diffused morning glow, and the shallow depth of field effect blurs the green forest in the background. 8K image quality, extremely rich in detail.
If you have an image of your choice, you can also pad the image to generate it.
Here's a shot of a girl taking a selfie Vlog while walking and talking, so it needs to be lip-synced.
Currently there are two mainstream methods of lip-synching, one is uploading pictures for lip-synching with Instant Dreams, and the other is uploading videos for lip-synching with Korin.
But that dream uploading pictures to lip-sync does not allow you to write cue words, and there is no way to control the actions of the protagonist, so here we use Korin to upload videos to lip-sync.
Gotta generate the video first.
We uploaded the images to Imagine Dreams or Korin for a graphic born video.
Even if you are a member of Instant Dreams, the images and videos generated on Instant Dreams can't eliminate the watermark in the upper left corner, so if you mind this watermark, it is recommended to generate images and videos on Kerin as shown in the picture below.
Based on the images in Subtitle I, the raw video prompt words are as follows:
Camera shaking, selfie view, camera following, walking in the direction of the camera while smiling and talking non-stop, the other hand is occasionally making some movements as if introducing something, starting at the fifth second the smile slowly disappears, the face reveals doubt, talking non-stop, looking left and right as if searching for some kind of sound
Here's an explanation of why the prompts are written the way they are.
That is, the model of Dreamborn video supports multi-camera generation since 3.0, and our purpose is to let the female protagonist hear the sound of Wang Wei's poem in the process of hiking, and she feels puzzled and then searches for the source of the sound. So we need to limit the seconds from the first to the second when the heroine is hiking normally, and then from the second to the second when she is puzzled to find the sound. The whole framework is like this, some of the details such as action ah expression ah, according to your needs to add their own.
The 10-second clip above is still complete enough to show the whole process of the heroine's movement and expression, so when we dub this 10-second video, we should try to dub a 9-second or 10-second voice. If the dubbing time is too short, the video will be cropped, and if the dubbing time is too long, it will be cut.
Now let's generate the voiceover first. Open Conch AI (https://www.minimaxi.com/audio), input the voiceover content, choose a tone you like, click Generate, and you get a 10-second sound.
I didn't set any parameters and was quite happy with the sound I got.
Next we're going to Korin (https://klingai.kuaishou.com) to dub the video.
In this screen of the raw video, we first click on this lip-sync on the left and then upload the video clip we generated earlier.
The interface will automatically jump to the window for editing audio, we upload the voiceover and click Add Voiceover.
At this time the audio will be automatically added to the lower left track, we can drag the audio or crop the audio to adjust, so that the audio appears in the video in the appropriate place, and finally click on Generate Now, probably wait a minute or two can be generated out.
lip sync
Here's also a side note on how to lip sync multiple people.
When we upload a video clip containing multiple people to Korin, Korin will recognize on its own that there are several characters.
As shown in the following figure, Kerin recognizes a male and a female, if we want to dub this female, we first need to select this female's avatar, and then upload this female's dub, which is to tell Kerin that we are uploading a dub for this female.
The uploaded voiceover for this female is going to be adjusted in the track area in the lower left corner, for example, if the example I gave is for this female to start talking from second 0 all the way through to second 3, then I'm going to place the material for this voiceover between seconds 0 and 3.
Then save the video when this voiceover is generated.
Next, we're going to dub the male voice in this same video. We'll upload the video we created from the female voiceover in the previous step, then we'll select the male avatar and upload the male voiceover, adjusting it from the 3rd to the 10th second, then we'll turn on the video soundtrack and click Generate Now.
When we dubbed this clip, the first time we dubbed it for a woman, the woman's dubbing took up seconds 0 to 3 of the video, so the man's dubbing could only be from seconds 3 to 10, and the second dubbing had to bring up the original sound of the video, that is, the woman's dubbing as well.
People may think AI is hard, so they stay put and don't want to try it.
However, I can tell you very responsibly that if you go for it, you can make it too!