Recreate the Viral Ink Groove AI Video with AnyCap and Cursor

Step-by-step tutorial to recreate the viral Ink Groove AI video. Full workflow using GPT Image 2, Seedance 2, AnyCap CLI, and Cursor with complete prompts and pro tips.

by AnyCap

The AI Video That Took Twitter by Storm

On May 6, 2026, designer and AI explorer Latte posted a 15-second video titled "Ink Groove" — and it exploded.

3,360 likes. 361 retweets. 41 quote tweets. The video shows a hip-hop dancer wielding an oversized ink brush in an infinite white void, with every swing leaving behind dense black ink trails. The camera choreography is stunning: low-angle tracking, rapid push-ins, spinning circular reveals, and a final explosive lunge that floods the entire frame in calligraphic black.

It was made entirely with GPT Image 2 and Seedance 2 in Dreamina, and the full workflow and prompts were shared publicly.

I wanted to see if the same result could be replicated without Dreamina — using only Cursor and AnyCap. And it worked. Here is the result:

No Dreamina. No SeaArt. No platform switching. Just the Cursor terminal and AnyCap CLI — one editor, one flow.

This guide is the exact process I followed. By the end, you will be able to produce the same result.


Why Cursor and AnyCap? The Workflow Nobody is Talking About

Most tutorials for GPT Image 2 and Seedance 2 assume you are working inside a web platform like SeaArt, Dreamina, or Venice.ai. That works — but it also means:

  • Switching browser tabs between image generation and video generation
  • Uploading and downloading files between platforms
  • Managing separate accounts and API keys for every tool
  • No version control or project-based organization

Cursor and AnyCap flips this entirely. AnyCap is a runtime capability layer that installs as a single CLI binary inside the Cursor integrated terminal. After one command:

npx -y skills add anycap-ai/anycap -a cursor -y

You get access to GPT Image 2 for storyboard generation and Seedance 2 for video production from the same terminal session. No browser tabs. No platform accounts. Just natural language commands that generate images and videos directly into your project workspace.

With AnyCap in Cursor, your entire AI video pipeline is:

  1. Write a prompt, get a storyboard image
  2. Feed that storyboard, get a video
  3. Iterate from the same terminal

Prerequisites: What You Need

  1. Cursor IDE installed
  2. AnyCap CLI installed inside the Cursor terminal:
curl -fsSL https://anycap.ai/install.sh | sh
anycap login
  1. Access to GPT Image 2 and Seedance 2 through AnyCap

That is it. Two tools. Get started with AnyCap — no platform accounts needed beyond your login.


Step 1: Generate a Director-Style PREVIS Storyboard

The secret to Ink Groove is the PREVIS storyboard — a director-style action storyboard that maps out every camera movement before any video is generated.

What is a PREVIS Storyboard?

PREVIS (pre-visualization) is a technique borrowed from film production. Instead of generating a single reference image and hoping the video model figures out the motion, you generate a storyboard that explicitly defines 8 continuous action frames showing movement trajectories, camera choreography, momentum trails, and ink flow patterns.

This dramatically reduces motion randomness — Seedance 2 has a clear visual roadmap to follow.

Use AnyCap GPT Image 2

Step 1: Generate a director-style PREVIS action storyboard in GPT using
simple natural language.

This helps reduce motion randomness and makes the action, camera movement,
and ink flow feel more intentionally directed.

Prompt:
"Create a director-style PREVIS action storyboard focused on 8 continuous
action frames using only minimal stick figures. No realistic anatomy,
clothing, pants, shoes, or character details. Show only movement, camera
choreography, momentum trails, and ink motion, with only a small amount of
necessary text."

The output is a grid of 8 stick-figure panels, each showing a different camera angle and movement phase. This becomes your reference image for the next step.


Step 2: Feed the Storyboard into Seedance 2

Now the magic happens. Take the PREVIS storyboard and feed it into Seedance 2 with a comprehensive video prompt that defines every aspect of the scene — subjects, environment, music, color palette, style, and a frame-by-frame timeline with camera choreography.

Use AnyCap Seedance 2

Step 2: Use the PREVIS board as the reference image for the Seedance prompt.

---

SUBJECTS
A hiphop dancer wearing a loose white top and dark oversized pants with
irregular ink erosion along the fabric edges; holding an oversized heavy
ink brush in the right hand with dense wet ink; the brush is always part
of the movement, generating visible ink trails and ink masses with every
swing.

ENVIRONMENT
Pure infinite white space with no visible floor separation.

MUSIC
minimal hiphop beat with clear groove and dynamic pacing;
0:00 to 0:08 steady rhythmic pulse with clean kick pattern;
0:08 to 0:11 slightly increased rhythmic density;
0:11 to 0:13 brief drop to near silence or minimal ambience;
0:13 to 0:15 strong low-frequency accent synchronized with final motion.

COLOR LOGIC
pure white background plus deep black and grayscale ink only, no additional
colors.

STYLE
ink wash minimalism plus abstract fashion silhouette plus dynamic motion
ink plus negative space composition plus ultra wide perspective distortion
plus clean high contrast rendering.

---

TIMELINE

The Timeline: Frame-by-Frame Director Cut

This is where the real magic lives. Each segment includes duration, camera movement, dancer choreography, and sound effects:

0:00 - 0:02
24mm low-angle close tracking shot. The dancer rapidly shifts weight in
groove motion, swinging the brush from behind the body to create the first
long ink arc while stepping sideways toward the camera; ink softly spreads
along the clothing edges.
SFX: (deep rhythmic bass, soft brush drag, flowing fabric motion)

0:02 - 0:04
The camera is pulled sideways by the arm motion, then snaps back forward.
Alternating arm swings and twisting steps create continuous hiphop rhythm;
thick crossing ink trails remain in the air, with some ink attaching to the
body as black structures.
SFX: (wide brush sweep, rhythmic cloth movement, airy motion texture)

0:04 - 0:06
The camera rapidly pulls back, then pushes in again. The dancer drops low
and explosively rises upward, swinging the brush vertically to create a
massive ink streak; the body half-rotates forward with strong wide-angle
distortion.
SFX: (upward brush motion, deep groove pulse, flowing air movement)

0:06 - 0:08
The camera slowly rotates while drifting backward. Continuous spinning
footwork and wide arm swings generate circular ink rings wrapping around
the dancer; body and ink structures constantly interweave.
SFX: (circular brush flow, layered fabric movement, soft rhythmic ambience)

0:08 - 0:10
The camera keeps retreating under pressure. The dancer aggressively advances
with three powerful steps, each paired with a large brush swing; layered
heavy ink structures rapidly stack, increasing visual density.
SFX: (dense groove rhythm, fast brush sweep, continuous motion texture)

0:10 - 0:12
The camera suddenly pushes close. Using forward momentum, the dancer
performs a large turning motion while the brush creates a full heavy ink
ring around the body; the figure rapidly passes the lens edge, briefly
breaking into ink strokes before reforming.
SFX: (wide sweeping motion, layered ink flow, rhythmic air movement)

0:12 - 0:13.5
The camera briefly stabilizes while still being pressured backward.
Maintaining groove momentum, the dancer pulls the brush far behind the body
into extreme tension, leaving a long dense ink tail while surrounding ink
gradually retracts inward.
SFX: (minimal bass pulse, stretched brush motion, soft ambient airflow)

0:13.5 - 0:15
Extreme frontal wide-perspective close shot. The dancer lunges forward with
full force, explosively whipping the brush toward the camera; massive
calligraphic ink strokes violently flood the frame until the entire screen
becomes pure black.
SFX: (deep bass accent, flowing ink spread, full frame brush motion)

This is what separates a director from a prompter: every second is accounted for, every camera angle justified, every sound cued.


Pro Tips for Matching the Original Quality

1. Camera Descriptions Are Everything

Vague direction like "the camera moves" produces random results. Specific direction like "24mm low-angle close tracking shot" tells Seedance 2 exactly what lens and motion to simulate. Use real cinematography terms: dolly, tracking, push-in, pull-back, pan, low-angle, wide-angle.

2. The PREVIS Storyboard Reduces Randomness

Without a storyboard, Seedance 2 guesses the sequence. With a PREVIS storyboard, it has a visual roadmap. Think of it as giving a choreographer sheet music instead of saying "dance."

3. Align Music Dynamics with Visual Pacing

The music prompt has a "drop to near silence" at 0:11 to 0:13, followed by a "strong low-frequency accent" at 0:13 to 0:15 — perfectly matching the final explosive lunge. Syncing audio and visual dynamics creates the cinematic impact that makes people hit share.

4. Start Small, Then Scale

Before generating the full 15-second video, test with shorter clips of 3 to 5 seconds to verify style and motion quality. Iterate on the prompt until the output matches your vision, then go for the full sequence.

5. Embrace the Constraints

"Pure black and white." "No visible floor separation." These are not limitations — they are what make the video distinctive. The most viral AI videos often come from the tightest creative constraints.


Why This Workflow Changes AI Video Production

From Prompter to Director

Most people use AI video tools like slot machines: write a prompt, pull the lever, see what comes out. The PREVIS and Seedance workflow turns you into a director. You plan shots. You choreograph camera movements. You sync audio to visuals. You think in timelines.

This is not about better prompts — it is about better process.

Cursor and AnyCap: The Unified Pipeline

Cursor gives you an editor-native workflow. AnyCap gives you multi-model access through a single CLI. Together, they eliminate the friction of platform-hopping:

  • Generate storyboard images: anycap image generate
  • Produce video: anycap video generate
  • Store outputs directly in your project workspace
  • Iterate: modify the prompt, regenerate, compare versions

No browser tabs. No upload and download loops. No context switching. Just a terminal and your creative direction.


FAQ

Can I use this workflow without Cursor?

Yes. AnyCap works as a standalone CLI on any system. Cursor just provides a smooth experience because the terminal is integrated into the editor.

What if my Seedance 2 output does not match the storyboard?

Start with shorter clips of 3 to 5 seconds. If the motion is still off, add more detail to specific timeline segments — particularly camera directions like "24mm low-angle" or "rapid push-in."

How much does this cost?

Generation costs vary by model and output length. Check AnyCap pricing for current rates on GPT Image 2 and Seedance 2.

Can I use this technique for other types of videos?

Absolutely. The PREVIS storyboard plus timeline approach works for any video that requires controlled choreography: product demos, animated shorts, dance sequences, action scenes, and more. Adapt the style keywords and subject descriptions to your use case.

Where can I see the original Ink Groove?

The original video and full prompt thread can be found on X — view the original creator thread with all prompts and community discussion.


Ready to try it yourself? Install AnyCap in Cursor, grab the prompts above, and start generating.