You’ve probably used AI to write captions, generate ideas, or maybe even create images.
But when it comes to video—you’re still stuck.
- You spend too long thinking of ideas
- Writing scripts feels slow
- Editing takes forever
And somehow, a 15-second video ends up costing you hours.
The issue isn’t your skill level.
It’s your process.
The fastest creators today aren’t just “using AI tools”—
they’re building AI workflows.
What you’re about to learn is a simple but powerful system:
👉 A 3-step workflow using Flowith + Seedance 2.0
👉 That turns ideas into finished videos in minutes
We’ll use a Hong Kong cha chaan teng ad example—but you can apply this to any business.
Step 1: Use Flowith to Generate Ad Concepts
Most people start by “thinking.”
That’s already inefficient.
Instead, you want AI to generate structured ideas for you—based on context.
Here’s a ready-to-use prompt you can paste directly into Flowith:
You are a Hong Kong-based advertising creative متخصص in short-form content (IG Reels, Threads) with deep understanding of local culture and Cantonese communication style.
Generate 3 different 15-second video ad concepts for the following brand:
[Brand Info]
Brand Name: Wah Kee Cha Chaan Teng
Location: Sai Yeung Choi Street, Mong Kok
Product: Lunch Set (Luncheon Meat Egg Rice + Iced Lemon Tea)
Price: HKD $42
Target Audience: Office workers and students nearby
Platform: IG Reels / Threads
[Creative Requirements]
- Must feel local, authentic, and relatable (Hong Kong style, not corporate)
- Must include natural Cantonese dialogue or voiceover
- Fast-paced and suitable for short-form video
- Scenes must be simple and easy to execute or generate
- Style can be humorous, emotional, or slice-of-life
[Output Format for each concept]
1. Concept Name
2. Core Hook (1 sentence that stops scrolling)
3. 15-second visual description (one continuous shot)
4. Cantonese dialogue (natural spoken tone)
5. Emotional tone
Ensure all 3 concepts are clearly different in style and angle.
The key here isn’t creativity—it’s structure and constraints.
The more specific your inputs, the better your outputs.
Step 2: Convert Concept into Seedance 2.0 Prompt
This is where most people fail.
They stop at ideas—but don’t translate them into visual instructions.
To generate video properly, you must describe:
- What’s happening
- How the camera moves
- What the environment looks like
- What language is spoken
Here’s a production-ready prompt you can use:
I have selected a Hong Kong cha chaan teng lunch advertisement concept.
Convert it into a highly detailed Seedance 2.0 video generation prompt.
Requirements:
- Duration: 15 seconds
- One continuous shot (no cuts)
- Natural Cantonese dialogue must be spoken by the character
Dialogue (must be included exactly):
"呢個套餐㗎，我自己都日日食㗎。四十二蚊啫，你試吓啦。華記冰室，等你嚟。"
Include:
Subject & Action:
- Middle-aged Hong Kong female boss
- Standing behind counter
- Holding or pointing to luncheon meat egg rice + iced lemon tea
Camera:
- Start close-up of food
- Slowly push in to subject
- Slight handheld movement
Environment:
- Authentic Hong Kong cha chaan teng
- Tiled walls, handwritten menu
- Red stools, busy lunch crowd
- Steam visible
Lighting:
- Fluorescent + warm tones
- Slight yellow-green grading
Style:
- Documentary, authentic, local Hong Kong vibe
Audio:
- Cantonese speaking
- Background kitchen noise
Keep it under 150 words, precise and visual.
Critical detail:
👉 Always explicitly state “Cantonese dialogue”
Otherwise, the model may default to English.
Step 3: Generate the Video with Seedance 2.0
You can use the output from Step 2.
Or just copy this optimized version directly:
A 15-second one-take documentary-style video inside an authentic Hong Kong cha chaan teng. The camera starts with a close-up of a steaming plate of luncheon meat and fried egg rice, with a glossy runny yolk, and an iced lemon tea beside it with visible condensation. The camera slowly pushes in toward a middle-aged Hong Kong female boss standing behind the counter. She looks directly into the camera and speaks naturally in Cantonese: "呢個套餐㗎，我自己都日日食㗎。四十二蚊啫，你試吓啦。華記冰室，等你嚟。" She gestures toward the food warmly. Background shows a busy lunch scene with customers eating, tiled walls, handwritten menu boards, red stools, and kitchen steam. Lighting is mixed fluorescent and warm tones, slightly yellow-green graded. Slight handheld movement, raw, authentic Hong Kong documentary style.
In about 30–60 seconds, you’ll get a complete video with:
- Cantonese voice
- Local environment
- Realistic motion
- Emotional tone
Here’s what really matters:
You’re not just learning how to “use AI tools.”
You’re learning how to build a content production system.
Once this workflow is in place, you can:
- Generate multiple ads per day
- Test different hooks quickly
- Scale content without burning time
And most importantly—
👉 Stop relying on inspiration
👉 Start relying on systems
That’s where real leverage comes from.
