Create an AI selfie vlog with Seedance 2.5

Step by step

Run the workflow

1. 01

   Prepare your face and voice references

   1. Prepare two clear face photos: one front-facing and one side-angle.
   2. Record about 30 seconds of your voice. Reuse these files for every scene.

   

   

2. 02

   Set up Seedance 2.5

   • Click here to open Higgsfield (opens in a new tab)The Reel was sponsored by Higgsfield.

   1. Open Video → Seedance 2.5. Choose Reference and upload your two photos and voice recording.
   2. Set 720p, 16:9 and audio on.
   3. Use the prompts below, one per scene. Change the appearance, outfit, location and dialogue for your own vlog.
3. 03

   Busy shopping street

   Set the duration to 5 seconds. Copy the prompt below and generate.

   Working asset

   Busy shopping street: Seedance 2.5

   Selfie-style vlog shot: he is filming himself at arm's length, the lens pointing back at his face. We see his face and his outstretched right arm reaching toward the lens from the frame edge; the hand at the end of that arm is always out of frame. His left hand is empty and free, swinging as he walks or pointing at the crowd. No device, screen or gadget visible anywhere in the image. Looks like real handheld footage: wide-angle, everything in focus, wobble, walking bounce, low-light noise, not cinematic, no shallow depth of field. He behaves like a real person vlogging for friends: talks while moving, glances between the lens and the street, natural expressions. He talks the way people actually talk on a vlog: fast, uneven, repeats words, reacts mid-sentence, never reciting. Night, walking through a busy Tokyo shopping street, people passing close. In the voice from the audio reference he says something like: "Okay... first night in Tokyo, bro, it's so packed. Like, look at this." The man is the person in the reference photos: round face, full cheeks, soft wide jaw, narrow eyes with heavy upper lids, low straight brows, wide nose base, full lips, curtain fringe, silver cross dangle earring, wearing a cream ribbed knit polo. Signs blurred, no logos, no cartoon characters. Street noise. One continuous shot, no cuts.

4. 04

   Ramen stall

   Set the duration to 5 seconds. Copy the prompt below and generate.

   Working asset

   Ramen stall: Seedance 2.5

   Selfie-style vlog shot: he is filming himself, holding the shot in his left hand just out of frame, the lens pointing back at his face from slightly above the bowl. The only hand visible is his right hand, holding wooden chopsticks. No device, screen or gadget visible anywhere in the image. Looks like real handheld footage: wide-angle, everything in focus, wobble, low-light noise, not cinematic, no shallow depth of field. He behaves like a real person vlogging for friends. He talks the way people actually talk: fast, uneven, repeats words, talks with his mouth full, never reciting. Night, at a tiny Tokyo ramen street stall, warm bare bulb, steam. Mid-meal: he slurps a big mouthful of noodles with the chopsticks, and while still chewing says into the lens, in the voice from the audio reference, something like: "Mm... look at the way I eat. Look! Look." Then he slurps again. The man is the person in the reference photos: round face, full cheeks, soft wide jaw, narrow eyes with heavy upper lids, low straight brows, wide nose base, full lips, curtain fringe, silver cross dangle earring, wearing a cream ribbed knit polo. No logos. Loud slurping, stall chatter. One continuous shot, no cuts.

5. 05

   Skin close-up

   Set the duration to 5 seconds. Copy the prompt below and generate.

   Working asset

   Skin close-up: Seedance 2.5

   Selfie-style vlog shot: he is filming himself, holding the shot in his right hand which stays out of frame, the lens pointing back at his face. His left hand is empty. No device, screen or gadget visible anywhere in the image. Looks like real handheld footage: wide-angle, everything in focus, wobble, low-light noise, not cinematic, no shallow depth of field. He behaves like a real person vlogging for friends, loose and natural. He talks the way people actually talk: fast, uneven, repeats words, reacts mid-sentence, never reciting. Night, outside the ramen stall on a quiet street under a warm street lamp, still chewing. The shot starts at arm's length on his face and shoulders, then in one casual move he brings the lens right up close to his face, tilts his head to show his cheek, touches his cheek with the fingers of his empty left hand, and says in the voice from the audio reference something like: "Wait... look at my skin texture. Look. Like... you can see everything." Up close: real pores, fine facial hair, a sheen of sweat, uneven tone, small blemishes. The man is the person in the reference photos: round face, full cheeks, soft wide jaw, narrow eyes with heavy upper lids, low straight brows, wide nose base, full lips, curtain fringe, silver cross dangle earring, wearing a cream ribbed knit polo. No beauty filter. One continuous shot, no cuts.

6. 06

   Overpass

   Set the duration to 4 seconds. Copy the prompt below and generate.

   Working asset

   Overpass: Seedance 2.5

   Casual selfie vlog filmed on his phone's front camera, the phone held in his own hand at arm's length, his arm reaching toward the camera from the frame edge; no other phone, no one else filming. Real phone video: wide-angle, everything in focus, handheld wobble, walking bounce, low-light noise, not cinematic, no shallow depth of field. He behaves like a real person vlogging for friends. He talks the way people actually talk: fast, excited, a bit disbelieving, repeats words, never reciting. Night, walking across a pedestrian overpass above a busy Tokyo road, car light trails below, wind in his hair. He leans into the phone and says in the voice from the audio reference something like: "Is this insanely real?? Isn't it?? Like be honest, can you even tell it's AI?" The man is the person in the reference photos: round face, full cheeks, soft wide jaw, narrow eyes with heavy upper lids, low straight brows, wide nose base, full lips, curtain fringe, silver cross dangle earring, wearing a cream ribbed knit polo. No logos. Traffic and wind noise. One continuous shot, no cuts.

7. 07

   Quiet lane

   Set the duration to 8 seconds. Copy the prompt below and generate.

   Working asset

   Quiet lane: Seedance 2.5

   Casual selfie vlog filmed on his phone's front camera, the phone held in his own hand at arm's length, his arm reaching toward the camera from the frame edge; no other phone, no one else filming. Real phone video: wide-angle, everything in focus, handheld wobble, walking bounce, low-light noise, not cinematic, no shallow depth of field. He behaves like a real person vlogging for friends, talking while walking, relaxed and natural, never reciting. Night, walking down a quiet narrow Tokyo residential lane with paper lanterns, potted plants and warm lit windows. In the voice from the audio reference he says, casually but clearly, this exact sentence in the middle: "Okay okay... comment AI to get the prompts plus seven days of unlimited access to Seedance two point five... I'll send you exactly how I made this." The man is the person in the reference photos: round face, full cheeks, soft wide jaw, narrow eyes with heavy upper lids, low straight brows, wide nose base, full lips, curtain fringe, silver cross dangle earring, wearing a cream ribbed knit polo. No vending machines, no logos. Quiet night ambience, footsteps. One continuous shot, no cuts.

8. 08

   Put your vlog together

   1. Arrange the clips in your video editor and trim them into one vlog.
   2. Keep the generated voice, add subtitles and export.