How to Use Seedance 2.0: Complete Tutorial 2026

TL;DR

Platform	Region	Free Tier	Best For
Dreamina (CapCut)	International	800s free / 30 days	Easiest access, full features
CapCut Desktop/Mobile	International	Included with CapCut	Edit + generate in one app
BytePlus ModelArk	International	8 free generations	Developer testing, API preview
Jimeng (即梦)	China	Free with queue	Full 2K, 15s, all features
Little Skylark (小云雀)	China (iOS)	1,200 credits + 120/day	Mobile-first workflow

Fastest way to start: Go to dreamina.capcut.com, sign in with Google, select Seedance 2.0, and generate your first video in under 2 minutes.

What Is Seedance 2.0?

Seedance 2.0 is ByteDance's latest AI video generation model, released February 2026. It generates 2K-resolution video (up to 15 seconds) from text, images, video, and audio inputs — simultaneously.

What makes it different from Sora, Kling, and Runway:

Quad-modal input: Upload up to 12 reference files (9 images, 3 videos, 3 audio clips) to guide generation
Native audio: Generates synchronized dialogue, sound effects, and music as part of the video — not stitched on afterward
Character consistency: Maintains identity across multiple shots
@ reference system: Tag specific files to control exactly how they influence the output

Where to Use Seedance 2.0

1. Dreamina (International — Recommended)

URL: dreamina.capcut.com

This is the easiest way to access Seedance 2.0 outside China. Dreamina is ByteDance's AI creative platform, now integrated under the CapCut brand.

Free tier: 800 seconds of free video generation, valid for 30 days from first sign-in. How to sign up:

Go to dreamina.capcut.com
Click "Sign In" → use Google, Apple, or email
You're in. No credit card required

How to generate a video:

Click "Video Generation" or "AI Video"
Select Seedance 2.0 from the model dropdown
Choose your mode:

- Text to Video — describe the scene in text - Image to Video — upload an image and describe the motion - Reference Mode — upload multiple reference files (images, videos, audio)

Set parameters:

- Aspect ratio: 16:9, 9:16, 4:3, 3:4, 1:1, or 21:9 - Duration: 5s, 8s, 10s, or 12s - Resolution: 720p (free) or 1080p (paid)

Write your prompt (see Prompt Framework below)
Click Generate

Paid plans:

Standard: ~$4.99/mo (more credits, 1080p)
Pro: ~$9.99/mo (2K resolution, priority generation)

2. CapCut (Desktop & Mobile)

URL: capcut.com

If you already use CapCut for editing, Seedance 2.0 is built right in.

Desktop:

Open CapCut desktop app
Create a new project
Go to Media → AI Media → AI Video
Select Seedance 2.0 model
Enter prompt or upload image
Generate → video appears directly in your timeline

Online:

Go to capcut.com → Online Editor
Open Script panel
Input your idea → Choose video length and style
Click Create — Seedance 2.0 generates the video
Edit in CapCut's timeline with transitions, captions, music

Mobile (iOS/Android):

Open CapCut mobile app
Navigate to AI Video tools
Select Seedance 2.0
Generate and edit on your phone

Free tier: Same as Dreamina — 800 seconds free for 30 days. Why use CapCut over Dreamina? If you need to edit the video afterward (add captions, transitions, music, effects), CapCut gives you a full editor alongside generation. Dreamina is better for pure generation.

3. BytePlus ModelArk (Developer Preview)

URL: byteplus.com/en/experience/modelark

BytePlus is ByteDance's enterprise cloud platform (the international version of Volcengine). Seedance 2.0 is available in the ModelArk playground.

Free tier: 8 free generations, no login required initially. How to use:

Go to the ModelArk playground
Select Seedance 2.0
Upload reference files or enter text prompt
Generate at up to 1080p, 15 seconds

API access: The official Seedance 2.0 API is expected to launch through BytePlus in late February 2026. Developers can currently debug in the BytePlus console. Best for: Developers who want API access or want to test before integrating programmatically.

4. Jimeng / 即梦 (China)

URL: jimeng.jianying.com

Jimeng (即梦) is the Chinese version of Dreamina. It's the most feature-complete platform for Seedance 2.0.

Requires: Douyin account (Chinese phone number needed) Advantages over international versions:

Full 2K resolution
Full 15-second generation
All features available on day one
Commercial licensing with paid plans

Free tier: Free with queue delays. Subscription (~68 RMB/mo) removes queue and unlocks commercial use. Limitation: Realistic human face uploads are prohibited due to Chinese regulations.

5. Little Skylark / 小云雀 (China, iOS)

A mobile app by ByteDance with generous free credits.

How to use:

Download "小云雀" (Little Skylark) from the Chinese App Store
Sign up → receive 1,200 credits + 120 daily credits
Go to video generation tab
Important: Manually select Seedance 2.0 from the model dropdown (it defaults to an older model)
Generate

Best for: Mobile-first users in China who want free credits.

The Prompt Framework

Seedance 2.0 responds best to structured, director-style prompts. Here's the framework:

The 5-Part Structure

Subject: [who/what — one person or object]
Action: [what they're doing — specific verb, present tense]
Camera: [shot size] + [movement] + [angle], [focal length]
Style: [one visual anchor], [lighting], [color treatment]
Constraints: [what to avoid], [duration], [frame rate]

Key rule: Shorter prompts with this structure outperform long, descriptive paragraphs. Aim for under 60 words plus constraints.

5 Copy-Paste Prompt Examples

1. Product showcase (clean & professional)

A sleek smartphone rotating slowly on a white marble surface.
Close-up, slow dolly-in, locked horizon, 50mm lens feel.
Soft key light with rim highlight, neutral color grade, light film grain.
No logos, no text overlays, no lens flares. Hold final frame 2s. 6-8s.

2. UGC-style (authentic feel)

A young woman unboxing a package at her kitchen table, smiling.
Medium shot, handheld phone perspective, slight natural sway.
Natural indoor window light, ungraded look, warm tones.
No captions, no snap zooms, no background music. 8-10s.

3. Cinematic landscape

A lone hiker walking along a mountain ridge at golden hour.
Wide establishing shot 2s, then slow push to medium, gimbal-smooth.
Overcast sky transitioning to warm sunset, muted blues and golds.
No Dutch angles, no crowds, no text. 10-12s.

4. Talking head (stable)

A man in his 30s speaking directly to camera in a modern office.
Medium close-up, locked tripod, subtle dolly-in over 10s.
Soft key light from 45 degrees, clean background separation, shallow depth.
No auto captions, no whip pans, no background movement. 12-15s.

5. Quick montage (social media)

Coffee being poured into a ceramic mug. Steam rising.
Beat 1: overhead close-up of pour (2s). Beat 2: side angle, mug on table (2s). Beat 3: hand wrapping around mug (2s).
Warm café lighting, analog film look.
No text overlays, no speed ramps. Cuts between beats. 6s total.

Prompt Tips That Actually Matter

Motion: Use one verb per shot. "Walking slowly toward camera" works. "Walking, turning, and waving" causes drift. Camera vocabulary that works:

Movement: dolly, track, crane, handheld, gimbal
Speed: pair with distance ("slow dolly-in, 1-2 feet")
Shot size: wide, medium, close-up (locks composition)
Angle: eye level, low angle, high angle
Lens: wide (24-28mm), normal (35-50mm), telephoto (85mm+)

Lighting: Be specific. "Dramatic side lighting" beats "good lighting." "Warm rim light from behind" beats "nice backlighting." Audio triggers: Include words like "reverb," "muffled," or "metallic clink" to activate native audio generation. For example: "the sharp click of a shutter" generates that specific sound. What to ban (3-5 items per prompt):

Visual noise: text overlays, watermarks, lens flares
Identity drift: extra characters, mirrors
Camera chaos: snap zooms, whip pans, Dutch angles
Body artifacts: extra fingers, deformed hands

When Generation Misses

Wrong framing, right action? Tighten the Camera section first
Motion feels off? Swap "handheld" ↔ "gimbal" and set explicit speed
Style drifts? Replace multiple adjectives with one strong anchor reference
Subject mutates? Simplify the descriptor, use fewer nouns
Artifacts repeat after 3 tries? Change shot plan or step back to a wider frame

The Reference System (Advanced)

Seedance 2.0's killer feature is the @ reference system. Instead of describing everything in text, you upload files and tag them with specific roles:

What you can upload (per generation):

Type	Max Count	Max Duration	Format
Images	9	—	JPG, PNG, WebP
Videos	3	15s total	MP4
Audio	3	15s total	MP3

How references work:

Image as style reference: Upload a photo → Seedance matches the color palette, lighting, and visual style
Image as character reference: Upload a face → Seedance maintains that character's identity across generations
Video as motion reference: Upload a clip → Seedance replicates the camera movement and action rhythm (not the content)
Audio as rhythm reference: Upload music → Seedance syncs visual transitions to the beat

Example workflow:

Upload a reference photo of your brand's visual style
Upload a short clip showing the camera movement you want (a slow pan, a zoom-in)
Upload your brand's audio jingle
Write: "A person using a laptop in a modern workspace, typing, smiling at screen"
Seedance generates a video with your brand's look, that camera movement, synced to your music

This is why Seedance 2.0 is particularly strong for template-based production — create once, generate variations at scale.

Pricing Comparison

Platform	Free Tier	Paid Plans	Best Value
Dreamina/CapCut	800s / 30 days	$4.99-$9.99/mo	Best for most users
BytePlus	8 generations	API pricing TBD	Developers
Jimeng	Free (queued)	~$9.60/mo (68 RMB)	China users
Little Skylark	1,200 + 120/day	N/A	Free mobile access

Cost per video (approximate):

5-second 480p: ~1 credit ($0.03-0.07)
5-second 1080p: ~5 credits ($0.15-0.35)
10-second 1080p: ~10 credits ($0.30-0.70)

Is it worth paying?

If you're generating 1-3 videos for a specific project, the free tier on Dreamina (800 seconds) is plenty. If you're doing regular content production, the ~$5-10/mo plans are reasonable compared to alternatives (Sora 2 costs ~$1.00 per 10s clip, Veo 3.1 costs ~$2.50).

Limitations to Know

Duration: Max 15 seconds per generation. For longer videos, you'll need to generate multiple clips and edit them together (CapCut helps here). Long-form coherence: Beyond 15 seconds, character consistency degrades. Multi-shot stories require careful reference management. Complex physics: Fluid dynamics, hair, and fabric are improved but still hallucinate in edge cases. Human faces (China platforms): Jimeng prohibits realistic human face uploads due to regulations. International platforms (Dreamina, CapCut) have fewer restrictions. API availability: The official API launches late February 2026. Until then, use the web platforms or third-party API providers.

Seedance 2.0 vs. Other Tools: Quick Decision Guide

Choose Seedance 2.0 when:

You have reference assets (images, videos, audio) you want to incorporate
You need template-based production (same style, different content)
You want native audio co-generation
You're creating ads, music videos, or branded content

Choose Sora 2 when:

You need the most realistic physics simulation
You're doing narrative storytelling with emotional depth
Budget isn't the primary concern

Choose Kling 3.0 when:

You need the fastest, cheapest generation
You're scaling social media content
Simple text-to-video is sufficient

Choose Y Build Demo Cut when:

You need a product demo video specifically
You want zero manual effort (fully automated)
You're launching on Product Hunt and need assets fast
Your product is a web app or SaaS

Each tool solves a different problem. Seedance 2.0 gives you creative control over cinematic content. Demo Cut gives you a finished product demo without touching a video editor.

Building a product and need a demo video for launch? Y Build Demo Cut auto-generates professional product demos — AI script, voiceover, transitions, and one-click Product Hunt publish. No video editing skills required. Try it free.

Sources: