Recently, AI's graphics technology has evolved dramatically, followed by Google Nano-Banana, followed byDream4.0, IT APPEARS THAT AI CONTENT PRODUCTION IS FLOURISHING AND THAT THE ACTUAL LANDING IS LESS THAN SATISFACTORY。
NO, I HAD A FRIEND COME TO ME LAST WEEKEND AND SAID HE WANTED TO DO AN MTV CASE WITH AIDigital HumanTools
Finds him some, gets together for a couple of days, heygen, silica intelligence, i.e., vEO3, i.e. dreamKelingNanoAI, Flying Shadow, Silicon, Mirror, Clip, Dreamface, Mirror..
THERE'S A LOT OF PRODUCTS, AND THE DOGS FOUND OUT, AND THERE'S NOT MUCH TO USE IN MTV
It's limited, short-time video generation, bad mouth. It's a lot of trouble
SOME OF THE GREAT GODS DO MTVS THAT LOOK GOOD, AND THEY TALK TO THEM IN PRIVATE, AND THEY REVEAL THEIR HEARTS, THEIR WORKS. AT THIS STAGE, THERE ARE OFTEN COUNTLESS “TICK CARDS” ATTEMPTED AND LENGTHY POST-PROCESSING; THERE ARE ALSO HEAD-DIGITAL VIDEO NUMBERS MADE BY THE SO-CALLED AI, WHICH ARE SEEN IN DOZENS OF VIDEOS, ALL OF WHICH ARE PICKY, ONE END-TO-MOUTH, LESS THAN 10 SECONDS IN TOTAL, AND IN THE MIDDLE ARE TRADITIONAL VIDEOS, WHICH ARE FALSE
But now the good thing is the process is running, and the next step is to optimize the tools
AND THEN I'M GOING TO TAKE YOU ALL TO TRY TO MAKE A DIGITAL MTV WITH AN AI TOOL CHAIN。
The core tool chain is simple:
- google imagen4 for image design and graphics
- Clint finished the silent video
- Mureka clone music
- the spirit, the dream, the dreamface achieves a music-to-mouth video
Step one, image design
LET'S START WITH THE AI. YOU CAN WRITE YOUR OWN DRAWING TIPS, OR YOU CAN HAVE AI DO IT FOR YOU
In "AI to be beautiful DJ," using gemini/kimi inverted pictures, got a pretty DJ tip
WE SEND IT TO GPT5 OR ANY AI, HINTING, "REPLACE THE SINGER WHO IS RECORDING THE SONG IN THE STUDIO, THE REST UNCHANGED."
GPT5 GIVES A VERY GRAPHIC NARRATIVE:
“A researching reporte ceturing a young Asian femile expert english in singing. She’s scientific studyheads, with her eyes closed, conserving interpretation as she records in to a high-quantitude microphone. She is presented in a dazzling fashion low V-neck drawings, accommodating her status presence, the return in general making a general sense, with a sense of justice
“A VIDEO OF A STUDIO SET SHOWING A YOUNG ASIAN FEMALE SINGER SINGING WITH PASSION. SHE WAS WEARING A PROFESSIONAL RECORDING STUDIO EARPIECE, EXPRESSING FEELINGS BEHIND HER EYES AND RECORDING HIGH-QUALITY MICROPHONES. SHE'S WEARING A BRIGHT RAINBOW LOW V-COLLAR DRESS THAT HIGHLIGHTS HER STAGE. THE BACKGROUND IS A MODERN RECORDING STUDIO SET-UP WITH SOUND-ABSORPTION FOAMS AND TACTICALLY DEPLOYED AUDIO EQUIPMENT, WHICH DISPLAYS A PROFESSIONAL RECORDING ENVIRONMENT.”
He gave a draft
BUT I DON'T LIKE GPT WITH THE DALLE STYLE
Google Nano-Banana. It's not bad, but it's slightly worse
nano Banana called and entered the hint into Google Nano-Banana, and in a few seconds, she got a beautiful picture of the woman, a puzzle, which one do you think you should choose
I picked a 12-point little sister. The little sister at the beginning of this article is a resource。
But it turns out that the 11 point little sister should be chosen because the 12-point little sister had a hand move, had a video for a maximum of 10 seconds, had to do a one-minute video, and the hand wasn't very well handled, but Nano-Banana did not have the precision, I went to the volcano engine, called 4.0, turned it into 4K。
UPLOAD A PICTURE, ENTER A HINT: CREATIVELY UPSCALE TO 16KRESOLUTION
Get a high-level map
Step 2: Video production
Why are you using Clint? It's not as good as a dream
enter klingai.kuaishou.com
Unlike a dream, the Spirit does not support a "one picture-to-mouth" only for video-to-mouth, but for uploading third-party video-to-mouth
In the clint, you have to draw a video first, and then use the mouth-to-mouth function in the video that you generate
Select 2.1 Master Edition, upload pictures of little sister, enter a hint: "A female singer, with all her focus on recording songs in the studio, singing with the music, swaying gently with the music."
HERE'S A 200-CENT SCORE, FUCKING EXPENSIVE
STEP THREE: AI WRITES SONGS
The AAI song can be written in Suno, or in Cullen Manway, by launching the global premier music reasoning model, Mureka O1, combined with the Mureka V6 model and helping to create music。
- The power of Mureka O1 is so powerful that it supports the creation of a single-key symphony, pure music, simple models and advanced models of songs that produce high-quality musical works with sound close to the real person。
Step 4: Oral treatment
1. Visible effects
On the right-hand side of the video just created
Because the original video is only 10 seconds, it only supports 10 seconds of clones in one time and is generated many times
Because it's only 10 seconds, so it's only 10 seconds to cut the music in, and then then it's integrated
Or take a frame for a new 10 seconds, but it turns out that the quality of the tail frame to generate a new video is going to decline, so I'm not using it on a large scale, and it's really too expensive. And it's working
2. The effect of a dream
(1) Video production
It's the same thing that dreams do, the same thing that you can do, the same thing that you can do, the same thing that you can do
But with video-to-mouth, only basic modes can be selected; with master models, only pictures can be used to transfer videos
Look at the results
(2) Tusheng Video
If you use a graphic video, the character moves very hard, even a little ghost
3,DreamFace Effect
A simpler alternative would be to use the DreamFace feature to upload pictures or silent videos with songs to get finished quickly. It's probably a bit of a downside, but it's more efficient
The mouth is a little off, the mic and the mouth overlaps are flawed, but the advantage is that it is efficient, and next time you keep your mouth shut。
Well, that's the case today。