The main thing to share with you today is thatAI VideoUnlimited Duration Technique (keep the characters moving in unison and coordinated silky smooth)
We know that at present basically all the AI platforms produce videos are 5S, 10S clips, so how do we break through the 10S or 20S, 30S, etc. so that the start has been able to unlimited length? Also keep the effect very silky smooth like a video generation.
The next step is to break it down in detail.
Step 1: Prepare a characterization
Here for the purpose of demonstration, we need a character role, this simple direct AI generation, if you can not write the prompt word then directly let the AI generation.
Open the beanbag and enter the requirement.
For example: please help to generate a group, realistic style of photography, for the depiction of an Asian woman frontal photo prompt words, although she is not dressed in a revealing, but everywhere exudes a seductive charm of women, the background for the light gray solid color background, the whole body photo, the ratio of the image is 9:16, 8 pictures!
And of course here are some cos girls for you to modify and use
Cosplay cute girl, round face, head wearing pink bow hair accessory, long hair tied in double ponytail, tutu skirt, white knee high socks, standing position fixed machine jumping, portrait, young girl facing the audience straight on, hands above her head, petite, background bokeh, full body photo.
Cosplay cute girl, round face, head wearing rabbit ears hair accessory, soft long hair tied in low ponytail, pleated skirt, pink stockings, standing position fixed camera rotating, portrait, young girl facing the audience straight on, hands caressing skirt, round body, background bokeh, full body photo.
Cosplay cute girl, doll face, head wearing bear hair accessory, long hair falling naturally, tutu skirt, white stockings, standing position fixed machine posing cute pose, portrait, young girl facing viewer straight, hands on both sides of cheeks, petite body, background bokeh, full body photo.
Cosplay cute girl, round face, head wearing wreath, long hair tied in high ponytail, dress, light color stocking, standing position fixed machine posing lively pose, portrait, young girl facing audience straight on, hands making heart gesture, well-proportioned body, background bokeh, full body photo.
Cosplay cute girl, round face, head wearing cat ears hair accessory, long hair tied in double ponytail, super short skirt, black pantyhose, standing position fixed machine dancing, portrait, girl facing audience straight, arms crossed, full figure, background bokeh, full body photo.
For the sake of demonstration, let's pick out the image that works well (preferably with a solid color background, as it's easier to get it in place when generating the action descriptors later on, and it won't interfere with the character because of the background elements)
Step 2: Design character action cues
Because one of the important core of the unlimited duration is to achieve the effect of silky smooth, a fragment must be interrelated, so you need to design a whole set of actions of the characters, respectively, with 5S a time period cue word, of course, 10S a group is also possible. Here we also have to go on to use the beanbag.
Enter the prompt words in the figure:
I want the girl in the picture to do a sexy dance, in 5 second segments, 6 segments which are silky smooth and connected. Please provide me with 6 5 second video cues.
Here are 6 5-second video cues:
The girl gently twisted her waist, her feet in a cat-walk posture, her eyes seductive, her fingers lightly traced the curves of her body, accompanied by rhythmic music.
Opening in a squatting position, she then slowly rose, her hips swinging rhythmically, her arms stretching gracefully, her hair moving as she did so.
With one hand on your waist and the other sliding down your neck, your body twists and turns in waves, with alternating light lifts of your legs to show off your supple lines.
After spinning around in place, stand on your tiptoes, bend your knees slightly, give your hips a quick little wiggle, and tease the camera with your eyes.
Bend your back slightly, tilt your head back, slide your hands forward from behind your back, and swing your hips dramatically from side to side to show your confidence and charisma.
In a semi-squat position, bounce your legs quickly and keep your upper body steady while making gentle dance movements with your hands in front of your chest, closing and framing.
Note here: As the cue word is written by AI, individual actions or descriptions can not be generated in the AI video, the use of can be generated when the spirit will be prompted by the system does not allow the generation. So our video cue word must be regular action.
If out of the prompt word is not satisfied can be generated multiple times to choose a relatively suitable can be.
Then we generate the video, here we recommend the use of the spirit of the 2.1 high-quality results are really good, the character movement finger movements are clear, each action is silky smooth and natural.
But I'm not a member here and I don't have enough points to do it all. Used Step Back That Dream 3.0 to test it out.
With the first clip video, we then need to export the last frame of the video, here we cut the projection to export.
Exporting a Still Screen Demo
Of course, we can also export the picture of high-definition repair, so that the effect is better. HD repair methods: that is, the dream screen, starflow ultra clear all kinds of AI repair is possible.
We then took the second image and continued to generate the clip video using the second video cue word.
The other 4 segments can be completed by following this procedure.
Finally a compilation of the 6 clip video with BGM and you're done.
That's what I'm sharing today.Everyone can get hands-on, easy to learn, hurry up and try it~!