Sep 25, 2026
AI Filmmaking
How to Create Your Own Song and Turn It Into an AI Music Video with Seedance 2.5
Write a song with Suno, design consistent characters, and generate a nostalgic early-2000s VHS music video with Seedance 2.5. Every character and scene prompt from the project, ready to copy.
Two men in dark suits singing at a studio microphone against an orange backdrop, captioned Suno v6 + Seedance 2.5

Creating a music video traditionally requires a lot of resources. You need a song, actors, locations, camera equipment, a production crew, and a proper shooting schedule. But with the development of AI tools, the process of creating a music video has changed completely.

In this tutorial, I will show you how I created my own song and transformed it into a complete AI-generated music video using tools like Suno and Seedance 2.5. I will walk you through my entire workflow, from creating the song, designing the characters, and generating cinematic scenes, to finally editing everything together into a finished music video.

I have also included the prompts that I used throughout this project below, so you can follow the same workflow and experiment with creating your own AI music video.

Creating the Song with Suno

The first step of creating any music video is having a song.

For this project, I started by creating a simple vocal idea and using Suno to help generate the final track. The purpose of the original recording was not to create a perfect vocal performance, but to give the AI a reference for the melody, rhythm, and overall feeling that I wanted.

After uploading the audio into Suno, I experimented with different music styles until I found something that matched the direction of the video. For this project, I wanted the song to feel bright, energetic, youthful, and slightly nostalgic, so I focused on a pop style with a strong and memorable melody.

Once I had the final version of the song, I exported the audio and used it as the foundation for the music video.

Designing the Characters Before Creating the Video

One of the most important steps in AI video generation is creating consistent characters.

Before generating a single clip, I made a four-panel element sheet for every character: a headless full-body front view, a side profile, a back view, and a medium close-up. Attaching these sheets to every video prompt is what keeps the same four friends recognizable from the street to the convenience store to the tent party.

Early 2000s Character Sheet

This one is me. Attach your own photo as the reference image, and the sheet keeps your face while dressing you in early-2000s streetwear.

Four-panel element sheet on light gray background. Use the face from the reference image exactly in the close-up and side views. Panel 1 (left): Full-body front view from neck down — no head. Oversized vintage band tee (faded graphic), cargo pants with multiple pockets, white platform sneakers, chain wallet. Standing neutral. Panel 2 (center-left): Full-body side profile. Same character, same outfit, looking camera-right. Panel 3 (center-right): Full-body back view. Same character, same outfit, standing neutral. Panel 4 (right): Medium close-up from mid-chest to top of head. Face, neck, shoulders visible. Show the band tee and snapback detail. All panels: light gray background, neutral studio lighting, consistent scale, cinematic photorealism with film grain and realistic fabric texture. Authentic early 2000s streetwear.
Four-panel early-2000s element sheet with a headless front view, side profile, back view, and close-up of a young man in a band tee, cargo pants, and a backwards snapback
Four-panel early-2000s element sheet with a headless front view, side profile, back view, and close-up of a young man in a band tee, cargo pants, and a backwards snapback

The Rest of the Group

The other three friends don't need a reference photo. The prompts describe each person and outfit from scratch. My first versions looked a little painted and had visible brand logos, so these final prompts ask for a natural film-camera look and no brand names.

Four-panel element sheet on light gray background. Black male, early 20s, Y2K athletic/hip-hop style. Panel 1 (left): Full-body front view from neck down — no head. Oversized athletic tee (faded graphic, no brand visible), baggy track pants with side stripe detail, fresh white sneakers, visible gold chain necklace. Standing neutral. Panel 2 (center-left): Full-body side profile. Same character, same outfit, looking camera-right. Panel 3 (center-right): Full-body back view. Same character, same outfit, standing neutral. Panel 4 (right): Medium close-up from mid-chest to top of head. Face, neck, shoulders visible. Show the athletic tee and gold chain detail. All panels: light gray background, neutral studio lighting, consistent scale. Natural photographic look — shot on film camera with soft studio lights, organic color palette, natural skin tones, realistic fabric texture and drape, subtle grain, matte finish. Not painted, not hyper-sharp. Feels like candid studio photography from the early 2000s. NO visible brand names or logos.
Four-panel element sheet on light gray background. White male, early 20s, Y2K skater/grunge style. Panel 1 (left): Full-body front view from neck down — no head. Oversized graphic tee (faded band graphic, no brand visible), baggy cargo pants with multiple pockets, worn white sneakers, chain wallet with visible clip. Standing neutral. Panel 2 (center-left): Full-body side profile. Same character, same outfit, looking camera-right. Panel 3 (center-right): Full-body back view. Same character, same outfit, standing neutral. Panel 4 (right): Medium close-up from mid-chest to top of head. Face, neck, shoulders visible. Show the graphic tee and black hair detail. All panels: light gray background, neutral studio lighting, consistent scale. Natural photographic look — shot on film camera with soft studio lights, organic color palette, natural skin tones, realistic fabric texture and drape, subtle grain, matte finish. Not painted, not hyper-sharp. Feels like candid studio photography from the early 2000s. NO visible brand names or logos.
Four-panel element sheet on light gray background. White female, early 20s, Y2K low-rise cargo style. Panel 1 (left): Full-body front view from neck down — no head. Fitted baby tee (pastel color or small graphic, no brand), low-rise cargo capris with embroidered details, white platform sneakers, small crossbody shoulder bag. Standing neutral. Panel 2 (center-left): Full-body side profile. Same character, same outfit, looking camera-right. Panel 3 (center-right): Full-body back view. Same character, same outfit, standing neutral. Panel 4 (right): Medium close-up from mid-chest to top of head. Face, neck, shoulders visible. Show the baby tee and blonde/light brown hair pulled back. All panels: light gray background, neutral studio lighting, consistent scale. Natural photographic look — shot on film camera with soft studio lights, organic color palette, natural skin tones, realistic fabric texture and drape, subtle grain, matte finish. Not painted, not hyper-sharp. Feels like candid studio photography from the early 2000s. NO visible brand names or logos.
Element sheet of a young man in an oversized faded tee, striped track pants, white sneakers, and a gold chain
Element sheet of a young man with shaggy black hair in a faded band tee, olive cargo pants, and worn sneakers
Element sheet of a young woman in a light blue baby tee, low-rise cargo capris, white sneakers, and a small crossbody bag
The other three members of the group, each on a four-panel element sheet.

Generating Cinematic Shots with Seedance 2.5

Once the characters and scenes were prepared, I moved into Seedance 2.5 to generate the video clips.

For each scene, I focused on writing detailed prompts that described the subject, action, environment, camera movement, and visual style. The more specific the instructions are, the easier it is for the AI to understand the intended result.

For this project, I wanted to create a nostalgic early-2000s music video feeling. Before generating anything, I gave the AI agent the look I wanted so it would keep every scene consistent:

so i wanna start generating my videos. lock in these aesthetics The overall vibe should feel nostalgic, youthful, raw, and casually cinematic. I want it to look like it was filmed on an old handheld camcorder in the late 1990s or early 2000s. Visual style: - handheld camera movement - VHS / camcorder look - soft image quality - slight motion blur - blown highlights - low-light noise and grain - washed colors with slightly warm tones - occasional focus hunting - light chromatic aberration - subtle scan lines or analog texture - imperfect framing - natural lighting - candid, unpolished energy Mood: - playful - youthful - fun - expressive - a little chaotic in a good way - nostalgic and stylish without looking too polished Framing and camera language: - mostly handheld shots - a mix of wide shots, medium shots, and close-ups - casual zooms - off-center framing at times - realistic movement like a friend filming with a camcorder - energetic but believable camera motion Important: - keep the VHS aesthetic consistent across all shots - keep the characters looking stylish and youthful - avoid modern ultra-sharp cinematic polish - avoid looking too clean, sterile, or commercial - keep it feeling authentic, natural, and era-appropriate - snappy cuts

I also added a subtle VHS-inspired style during the editing process. The goal was not to heavily distort the footage with effects, but to add small imperfections that make the AI-generated scenes feel like they were captured with a real camera.

Small details such as film texture, handheld movement, and natural color adjustments can make a big difference in making AI footage feel more authentic.

Every clip below was generated with Seedance 2.5 at 16:9 and 480p. Attach the four character sheets as reference images 1 to 4 for every scene except the b-boy clip, which uses no references.

Skating Down the Street

Handheld camcorder footage, Y2K VHS aesthetic. Group of four young people from reference image 1, reference image 2, reference image 3, reference image 4 skateboarding down an empty suburban street toward the golden sunset. Wide handheld shot tracking their movement, slight motion blur, warm washed colors. Soft focus moment as they pass. Blown-out sky. Film grain and noise. Occasional focus hunting. Snappy cut at the end. Playful energy, nostalgic and raw. Natural lighting. 10 seconds.
The group skating toward the sunset. Seedance 2.5, 16:9, 480p, 10 seconds.
The group skating toward the sunset. Seedance 2.5, 16:9, 480p, 10 seconds.
0:00/0:00
The group skating toward the sunset. Seedance 2.5, 16:9, 480p, 10 seconds.

Convenience Store Hangout

Handheld camcorder footage, Y2K VHS aesthetic. Interior of brightly lit convenience store. Four young people from reference image 1, reference image 2, reference image 3, reference image 4 browsing shelves, grabbing snacks, joking around. Handheld camera moving between them. High contrast from fluorescent lights, blown highlights, washed colors. Soft focus moments. Low-light noise. Chromatic aberration visible. Casual framing, slightly off-center. Grainy texture. Candid, playful energy. 10 seconds.
Grabbing snacks under fluorescent lights. Seedance 2.5, 16:9, 480p, 10 seconds.
Grabbing snacks under fluorescent lights. Seedance 2.5, 16:9, 480p, 10 seconds.
0:00/0:00
Grabbing snacks under fluorescent lights. Seedance 2.5, 16:9, 480p, 10 seconds.

Parking Lot Sprint

Short clips like this one make good cutaways between longer scenes. The prompt asks for about two seconds of action, and I trimmed the best moment in the edit.

Handheld camcorder footage, Y2K VHS aesthetic. Four tall people from reference image 1 (1.8m+ height), reference image 2, reference image 3, reference image 4 sprinting through a parking lot, camera handheld and shaky, catching them mid-run. Warm sunset light, motion blur, slightly off-center framing, washed colors. Film grain, low-light noise, focus hunting. Playful, raw kinetic energy. ~2 seconds.
Sprinting across an open lot at sunset. Seedance 2.5, 16:9, 480p, 5 seconds.
Sprinting across an open lot at sunset. Seedance 2.5, 16:9, 480p, 5 seconds.
0:00/0:00
Sprinting across an open lot at sunset. Seedance 2.5, 16:9, 480p, 5 seconds.

Playing with a Flip Phone

Late 1990s/early 2000s VHS handheld aesthetic. Characters: two friends from reference image 1 and reference image 2 with reference image 3 and reference image 4 nearby. Scene: sitting on outdoor school steps after class, shoulders touching, scrolling through tiny flip phone together. One types slowly on physical keypad; other laughs at screen. Camera: medium three-quarter shot → pan down to phone → snap back to their reaction. Slight VHS focus hunting. Style: soft image, motion blur, blown highlights, grain, warm washed colors, chromatic aberration, scan lines.
Two friends laughing at a flip phone on the school steps. Seedance 2.5, 16:9, 480p, 10 seconds.
Two friends laughing at a flip phone on the school steps. Seedance 2.5, 16:9, 480p, 10 seconds.
0:00/0:00
Two friends laughing at a flip phone on the school steps. Seedance 2.5, 16:9, 480p, 10 seconds.

Room Party

Late 1990s/early 2000s VHS handheld aesthetic. Characters: four friends from reference image 1, reference image 2, reference image 3, reference image 4. Scene: playing music loudly from stereo, casually dancing inside cramped bedroom. Nothing choreographed—jumping in place slightly, pointing at each other, laughing, lip-syncing into hairbrush. Camera: energetic handheld medium framing, three-quarter angles, fast pans, abrupt zooms, occasional tilted framing. Style: soft image, motion blur, blown highlights, grain, warm washed colors, focus hunting, chromatic aberration, scan lines, fast snappy cuts.
Lip-syncing into a hairbrush in a cramped bedroom. Seedance 2.5, 16:9, 480p, 10 seconds.
Lip-syncing into a hairbrush in a cramped bedroom. Seedance 2.5, 16:9, 480p, 10 seconds.
0:00/0:00
Lip-syncing into a hairbrush in a cramped bedroom. Seedance 2.5, 16:9, 480p, 10 seconds.

Locker Hangout

Late 1990s/early 2000s VHS handheld aesthetic. Characters: group from reference image 1, reference image 2, reference image 3, reference image 4. Scene: casually hanging around lockers after school, leaning against wall, comparing phones, exchanging CDs, talking. Someone walks past camera briefly blocking frame. Camera: handheld medium-wide shot, sideways pan across group, three-quarter angles, quick reframing between conversations. Style: soft image, motion blur, blown highlights, grain, warm washed colors, focus hunting, chromatic aberration, scan lines.
Hanging around the lockers after school. Seedance 2.5, 16:9, 480p, 10 seconds.
Hanging around the lockers after school. Seedance 2.5, 16:9, 480p, 10 seconds.
0:00/0:00
Hanging around the lockers after school. Seedance 2.5, 16:9, 480p, 10 seconds.

Tent Party (Slow Shutter POV)

For the second half of the video, the group moves to a tent party. These prompts bring in a crowd of extras around the four main characters. This clip has the heavy, smeared motion blur of a slow shutter and a camera that feels like it's being carried through the crowd.

Late 1990s/early 2000s VHS handheld aesthetic. Characters: large crowd including main group from reference image 1, reference image 2, reference image 3, reference image 4. Scene: chaos of extras and main characters dancing wildly inside packed tent, shuffling, jumping, moving in all directions, some with raised hands, boombox visible in corner playing music loud. Warm tent lighting. Camera: very handheld and shaky, quick whip pans following dancers, sudden zooms on shuffle footwork and raised hands, tilted framing, following movement chaotically. Style: soft image, motion blur, blown highlights, grain, warm washed colors, focus hunting, chromatic aberration, scan lines, extremely snappy cuts.
The packed tent party, with slow-shutter blur and whip pans. Seedance 2.5, 16:9, 480p, 10 seconds.
The packed tent party, with slow-shutter blur and whip pans. Seedance 2.5, 16:9, 480p, 10 seconds.
0:00/0:00
The packed tent party, with slow-shutter blur and whip pans. Seedance 2.5, 16:9, 480p, 10 seconds.

Shuffle

Late 1990s/early 2000s VHS handheld aesthetic. Characters: main group from reference image 1, reference image 2, reference image 3, reference image 4 dancing with crowd of extras. Scene: people doing high-energy shuffling dance moves inside main tent, feet moving fast, energetic foot work, bodies swaying rhythmically. Hands up, jumping between shuffles. Fairy lights casting warm glow. Camera: shaky handheld medium shot, panning down to feet showing shuffle movement, quick whip pan to faces, snappy cuts. Style: soft image, motion blur, blown highlights, grain, warm washed colors, focus hunting, chromatic aberration, scan lines.
The group shuffling with the crowd under fairy lights. Seedance 2.5, 16:9, 480p, 10 seconds.
The group shuffling with the crowd under fairy lights. Seedance 2.5, 16:9, 480p, 10 seconds.
0:00/0:00
The group shuffling with the crowd under fairy lights. Seedance 2.5, 16:9, 480p, 10 seconds.

B-Boy Breakdancing

This clip doesn't use the character sheets at all. A random performer and crowd are exactly what a party cutaway needs.

Late 1990s/early 2000s VHS handheld aesthetic. Characters: different b-boy breakdancer, crowd watching. Scene: wide shot of b-boy doing windmill spins, flairs, and power moves in open space inside tent, crowd gathered around watching. High-energy athletic movement, body control, complex footwork patterns. Fairy lights visible overhead. Camera: handheld wide framing showing full body, quick cuts between different move sequences, sudden zooms on spinning motions, abrupt whip pans following the dancer, no slow movements. Style: soft image, motion blur, blown highlights, grain, warm washed colors, focus hunting, chromatic aberration, scan lines, rapid frenetic cuts.
A b-boy spinning in the middle of the crowd. Seedance 2.5, 16:9, 480p, 6 seconds, no reference images.
A b-boy spinning in the middle of the crowd. Seedance 2.5, 16:9, 480p, 6 seconds, no reference images.
0:00/0:00
A b-boy spinning in the middle of the crowd. Seedance 2.5, 16:9, 480p, 6 seconds, no reference images.

Putting It All Together

Creating an AI music video is not simply about generating one prompt and getting a finished result. The biggest difference comes from the creative process behind it.

The workflow is similar to traditional filmmaking: create the concept, design the characters, plan the scenes, generate the footage, select the best shots, and edit everything together.

With tools like Suno and Seedance 2.5, creators now have the ability to produce music videos and cinematic scenes that would previously require a much larger production budget.

Try the AI agent now and start creating your own music video!