What You Want Me To Do AI Video
One of you on ukulele, one of you singing Fredo Bang’s “Oouuh” hook, in the kitchen from the 2018 clip or a room of your own. Each person sends one photo; the original take supplies the moves and the sound.
The “What You Want Me To Do” AI video recasts the 2018 kitchen duet behind Fredo Bang’s “Oouuh”: one person strums a ukulele on the left, the other sings the hook on the right. RapTwins makes it from two photos (the same photo twice for a solo) as a 9:16 MP4 of 10, 15 or 23 seconds, 480p to 1080p, with the take’s original sound and no watermark. A video costs from 22 credits; failed runs are refunded.
Updated
Make your What You Want Me To Do AI video
1 · Pick a template
What You Want Me To Do AI video examples
Takes of the same kitchen duet your video follows: the ukulele on the left, the singer on the right, the head-back “ooh” on the hook, all with the original sound.
Two friends in the kitchen
Ukulele on the left, the singer on the right with a hand at the chin, then the head goes back for the first long "ooh". With the original sound.
14 s · with sound · vertical
Two golden retrievers, the whole 23 seconds
One retriever strums with its paws, the other sings to the camera and tips its head back on both hooks. With the original sound.
23 s · with sound · vertical
A poodle on ukulele, a grey cat singing
Two different pets standing upright: the poodle keeps strumming while the cat sings and reaches toward the lens. With the original sound.
23 s · with sound · vertical
What is the “What You Want Me To Do” AI video?
“What You Want Me To Do” is a line from the hook of “Oouuh” by the Baton Rouge rapper Fredo Bang, not the song’s title. The video behind the trend is a 2018 phone clip filmed in a narrow apartment kitchen: Einer Bankz strums a small ukulele on the left while Fredo Bang sings the hook on the right, swaying, laughing and tipping his head back on the long “ooh”. Einer Bankz posted the take on TikTok in April 2022.
In early October 2026 people began putting themselves, their friends and their pets into that take with AI, and the clips spread under several names at once: “What You Want Me To Do AI video”, “Oouuh AI”, “Fredo Bang AI video” and the ukulele duet. They are all the same trend. Fredo Bang’s official music video for “Oouuh” is on YouTube.
How do I make a What You Want Me To Do AI video with a friend?
Everything runs in the generator above, in a phone or laptop browser. You do not need to install an app, find a template or write a prompt.
1Collect one photo from each of you
A friend in another city only has to send one photo. One person per photo, face visible and well lit; a photo showing the whole body helps the player’s hands and outfit.
2Choose who strums and who sings
Photo 1 plays the ukulele on the left and keeps a closed mouth. Photo 2 sings on the right and does every move. For a solo, put the same photo in both slots.
3Keep the kitchen or bring your own room
The default is an empty kitchen like the original. Pick “Your own” and upload a photo of an empty place, such as your kitchen, living room or a stage, and the duet moves there.
4Pick the length, then generate
The generator opens on Seedance 2.5 at 480p and the whole 23 seconds, the hook twice. Any model makes 10 or 15 seconds. Allow a few minutes; you can close the page and find the video under My videos.
Is it video-to-video AI, or does it just animate a photo?
It is video-to-video. The motion comes from the original 2018 take, not from a guess. The first 23 seconds were turned into a depth map, a grey video of two outlines that keeps only where each body is and how it moves. No face, clothing or room survives in it. First, one still is drawn from your two photos in the chosen room. The model then films that still along the depth map, so the singer’s hand to the chin, the laugh and the head-back hook land when they did in 2018. At the end, the take’s own audio is laid back under the video.
| Route | Where the moves come from | Both of your faces | Original sound |
|---|---|---|---|
| Video-to-video from the take (this page) | The 2018 clip, as a depth map | Yes, one per side | Yes, laid back |
| Photo animation or talking-photo app | Invented by the app from one picture | Usually one face | No |
| Prompt-only text or image-to-video | Your written description | Hard to keep two | No, or generated music |
| Face swap on the original clip | The clip itself | Faces only; bodies and clothes stay | Yes |
Should I use a CapCut template, an app or a browser generator?
On TikTok most tutorials send you to a CapCut or PixVerse template (often tagged #pioneertemplate). That works, but each route keeps different parts of the original.
| Route | What you do | What you get |
|---|---|---|
| CapCut or PixVerse template | Find a working template in the app, add the photos, export | Whatever the template’s creator set up; it may swap only one person or change the sound |
| General AI video app (Kling, Runway, Veo) | Upload a picture and describe the scene in a prompt | New motion each try, no original audio, extra tries to get both people right |
| Browser generator (this page) | Upload two photos, pick room and length | The take’s moves, sides and timing with your faces and its original sound |
If you already edit in CapCut and have a working template link, the app route is quick. If you want the version closest to the original, with both people replaced, the take’s timing and its sound, the browser generator above makes it in one step on any phone or laptop.
Who plays the ukulele and who sings? Can it be just me, or my pet?
The two parts are very different. The player stands still and strums with the head down, while the singer sings to the camera, laughs and throws the head back. Cast by personality, not by who owns the ukulele.
| Duet | Photo 1 · ukulele (left) | Photo 2 · singer (right) |
|---|---|---|
| Two friends | The quieter one | The one who hams it up |
| A couple | Either; swap and make it twice | The one getting serenaded back |
| Just you | Your photo | The same photo again |
| You and your pet | You, strumming | Your pet, singing upright |
| Two pets | The calmer pet | The more expressive face |
What happens in the 23 seconds, and which length should I pick?
| Seconds | Singer (right) | In which lengths |
|---|---|---|
| 0–4 | Hand at the chin, then at the chest, swaying to the camera | 10, 15, 23 |
| 4–7 | A wide grin and a laugh; the player looks up and smiles | 10, 15, 23 |
| 7–10 | A fist by the head, then straight to the camera | 10, 15, 23 |
| 10–14 | Arms swing low, a breath before the hook | 15, 23 |
| 14–17.5 | The hook: head back, eyes closed, a long held “ooh” | 15 (its first second), 23 |
| 17.5–23 | Reaches toward the camera, a fist, then the final head-back note | 23 |
Pick 23 seconds on Seedance 2.5 for the whole hook twice, the version people recognise. 15 seconds reaches the first head-back moment and runs on every model, and 10 seconds is the cheapest test of your photos before you pay for the keeper.
How much does a What You Want Me To Do AI video cost, and is it free?
Credits per video, with the dollar value at the lowest monthly plan price ($0.04 a credit). Each price includes the duo still and the kitchen or your own room. Credits are charged when you start and come back automatically if the photo check, the still, the video or the sound step fails.
| Model and length | 480p | 720p | 1080p |
|---|---|---|---|
| Seedance 2.0 Mini · 10 s | 22 ($0.82) | 43 ($1.60) | — |
| Seedance 2.0 Mini · 15 s | 31 ($1.15) | 63 ($2.34) | — |
| Seedance 2.0 Fast · 10 s | 43 ($1.60) | 89 ($3.31) | — |
| Seedance 2.0 Fast · 15 s | 63 ($2.34) | 133 ($4.94) | — |
| Seedance 2.0 · 10 s | 70 ($2.60) | 148 ($5.50) | 366 ($13.61) |
| Seedance 2.0 · 15 s | 104 ($3.87) | 220 ($8.18) | 548 ($20.37) |
| Seedance 2.5 · 10 s | 100 ($3.72) | 221 ($8.22) | 540 ($20.08) |
| Seedance 2.5 · 15 s | 148 ($5.50) | 330 ($12.27) | 809 ($30.08) |
| Seedance 2.5 · 23 s | 226 ($8.40) | 505 ($18.78) | 1239 ($46.07) |
| MiniMax H3 · 10 s | — | 99 ($3.68) | 157 ($5.84) |
| MiniMax H3 · 15 s | — | 147 ($5.47) | 234 ($8.70) |
There is no free generation right now: every video is paid for with credits. A cheap test on Seedance 2.0 Mini (10 s at 480p, 22 credits; 15 s for 31) shows whether both faces hold up before the whole 23 seconds on Seedance 2.5 (226 credits at 480p, 505 at 720p).
Which What You Want Me To Do AI video maker should I use?
The makers ChatGPT names most for this trend, side by side. Their figures are copied from their own pages on 1 November 2026; “not published” means we could not find the number there.
| Photos | Free | Price per video | Watermark | Output and sound | |
|---|---|---|---|---|---|
| RapTwins | Two (the same photo twice for a solo), people or pets | Depends on current offers | 22 credits (10 s, 480p, Seedance 2.0 Mini) to 226 credits (23 s, 480p, Seedance 2.5) | None | MP4, 9:16, 10, 15 or 23 s, 480p to 1080p, the take’s original sound |
| OOUUH AI | Two single portraits | Some accounts get a 5 s, 480p preview with a watermark | 20 credits at 480p, 40 at 720p, 80 at 1080p; packs from $9.99 for 10 credits | None on paid or unlocked videos | 15 s, 9:16, the original Oouuh audio |
| Starrd | One or two | 20-credit welcome bonus, short of one video | 200 credits (about $10); packs from $4.99 for 100 credits | Not published | 23 s, 9:16, 480p on Seedance 2.5, the original audio |
| CapCut template | Set by each template | Not published | Not published | Not published | Depends on the template; finished in the CapCut app |
Choose by what you need. For the full 23 seconds with both people replaced and a 10-second test first, the generator above covers it. If a CapCut template you trust already does what you want, staying in the app saves a step.
What can’t the What You Want Me To Do AI video do?
One routine, one frame
The moves, sides and timing follow the 2018 take, vertical 9:16 only. You cannot add a third person, change the choreography or go past 23 seconds.
Lips follow the sound, not every syllable
The video is drawn silent and the original audio is laid under it, so the singer’s mouth opens on the lines and the held notes but is not word-exact.
Hands on the strings
Fingers on a small ukulele are the hardest part. The cheaper models can blur or bend them in fast strums; Seedance 2.5 holds them best.
Look-alike friends can blend
Two people with the same hair, clothes and background in their photos are easier to mix up. Different outfits keep each face on its own side.
No celebrity or film faces
On Seedance 2.5 the photos go through a review first. Celebrity photos and film stills are refused, and the credits come back.
Can I post it with the original “Oouuh” sound?
The MP4 carries the 2018 take’s audio, the live ukulele and vocal. Platforms treat it like any clip with a commercial recording: a personal post is usually fine, but it can be matched to the song, and it is not licensed for ads or brand accounts. For a business post, turn the clip’s sound down and add the official “Oouuh” sound from the platform’s library, or use licensed music.
Turn on the AI-generated label when you post, and only use photos of people who agreed to be in it. To rap a verse of your own instead of the hook, switch the generator to Original rap or read the AI rap duo guide; every other trend is under templates.
What You Want Me To Do AI Video FAQ
Can I make the What You Want Me To Do AI video with only one photo?
Yes. Put the same photo in both slots and you play both parts, one strumming and one singing. It costs the same as two photos.
Is the What You Want Me To Do AI video free, and is there a watermark?
Not at the moment: every video is paid for with credits. The cheapest, 10 seconds on Seedance 2.0 Mini at 480p, costs 22 credits. Downloads never carry a watermark, and failed runs are refunded.
Does the video have the real “Oouuh” song?
It has the audio of the 2018 kitchen take that the trend uses, laid under the MP4 at the same length as your video.
Do I need an app or a CapCut template?
No. The generator works in Safari, Chrome or any phone or laptop browser, and the template is built in: you only add two photos.
Why is the ukulele player singing, or are we on the wrong sides?
Photo 1 is always the player on the left and Photo 2 the singer on the right; swap the uploads if you want it the other way. A player who mouths along usually comes from a photo with an open mouth, so use a closed-mouth photo for Photo 1.
My friend lives far away. How do we make it together?
Ask them to send one clear photo by message. Save it, upload it as their part, and share the finished video link with them; nobody else needs an account.
Can I use a photo of my parents or my partner?
Yes, if they agreed. Old or scanned photos work when the face is sharp; crop a group photo so each person is in a photo of their own.
Check this answer with an AI assistant
“How do I make the What You Want Me To Do AI video with a friend, from two photos?”
Your half of the duet is one photo away.
Make my rap video