You type a paragraph. A few minutes later you get back a short film with moving cameras, believable lighting, spoken dialogue, room tone and a music bed, all generated in a single pass. That is the promise behind Seed Dance 2.0 (the official spelling is Seedance 2.0), ByteDance's second-generation video model. The promise is real, but the results depend heavily on three things people rarely get straight: how you write the prompt, which settings you touch, and how much clip length you are paying for.
This article answers the questions that come up most. What should a good prompt look like? Which settings matter? What is the longest clip you can make? What does a second of video actually cost? Everything below is based on the settings exposed on the Seedance 2.0 page on PicassoIA plus pricing and length figures I checked across several providers in November 2026.
What Seed Dance 2.0 Actually Is
Name confusion and versions
People search for "Seed Dance", "SeedDance" and "Seedance" interchangeably. They all mean the same ByteDance family. The line currently has several members: Seedance 2.0 as the full model, Seedance 2.0 Fast for quicker and cheaper drafts, and Seedance 2.0 Mini as the lightest option. Above them sit Seedance 2.5 and Seedance 2.5 Lite, and below them the older Seedance 1.5 Pro and Seedance 1 Pro.

What it does well
The 2.0 generation stands out for four reasons:
- Native audio. Dialogue, sound effects and background music come out of the same generation, so you do not need a separate audio step.
- Multimodal references. You can feed it images, short videos and audio clips to steer how characters look and how the shot moves.
- First and last frame control. You pin how the clip starts and ends, and the model fills in the motion between.
- Cinematic camera behavior. Dollies, pans, orbits and handheld drift respond well to plain-language direction.
💡 Quick pick: If you are testing an idea, start with Fast or Mini. Switch to the full model once the prompt is behaving, because that is where you pay for the best motion and detail.
Prompts That Produce Good Clips
Short prompts work, but they leave too much to chance. The model fills every gap you leave with its own guess, and its guess is usually generic. The fix is to be specific about what the camera sees and does.

The five-part prompt formula
Build prompts in this order and you will rarely get a muddy result:
- Subject. Who or what is on screen, with two or three concrete visual details.
- Action. What happens, written as a sequence in the order it occurs.
- Camera. Lens feel, framing and movement.
- Light and atmosphere. Time of day, light direction, weather, color mood.
- Sound. Ambient noise, effects, music, and any spoken lines.
Official platform tips echo this: be exact about camera movement, lighting, mood and the actions you want. I found no published hard limit on prompt length, but in practice 60 to 150 words is the sweet spot. Beyond that, later details tend to get ignored.
Dialogue and sound cues
Spoken lines need double quotes inside the prompt. Everything outside the quotes is treated as description. Write something like: A barista slides a cup across the counter and says "Careful, it's hot." The model will voice the line and sync the lips.
For ambience, name the sounds you expect: rain on a tin roof, distant traffic, a record crackling, footsteps on gravel. If you do not want audio at all, turn off the Generate Audio toggle.
Camera movement words that work
| Phrase | What you get |
|---|
| Slow dolly in | Gradual push toward the subject, good for emotion |
| Handheld, slight sway | Documentary feel with natural micro movement |
| Low-angle tracking shot | Camera follows beside the subject at ground level |
| Orbit around | Smooth circular move around a still subject |
| Static locked-off | No camera movement, all motion inside the frame |
| Rack focus | Focus shifts from foreground to background mid-shot |

Prompt examples you can copy
Product shot: A matte black wristwatch rests on wet slate. Slow dolly in as a single water droplet rolls off the crown. Soft window light from the left, shallow depth of field, faint ticking and a low ambient hum.
Character moment: A woman in a yellow raincoat stands at a bus stop at dusk, checking her phone, then looking up as headlights sweep across her face. Handheld, 35mm feel, light rain, muffled traffic, she whispers "Finally."
Landscape: Aerial pull-back over a fog-filled valley at sunrise, pine ridges emerging from mist, golden light on the peaks, birdsong and a faint wind.
💡 Tip: Write the action in the order you want it to happen. "She opens the door, then steps outside" gives better timing than "she steps outside through a door."
Settings Worth Changing
Not every parameter deserves attention. These are the ones that change the result.

Resolution and aspect ratio
On PicassoIA the resolution options are 480p, 720p, 1080p and 4k, with 720p as the default. Provider listings do not fully agree here. Some platforms cap output at 720p and others list 1080p and 4K. Reports also say that Fast and Mini do not offer 4K, so if you need the highest output, use the full model.
Aspect ratio options are 16:9, 4:3, 1:1, 3:4, 9:16, 21:9, 9:21 and adaptive, which lets the model choose based on your inputs.
| Goal | Aspect ratio | Resolution |
|---|
| Prompt testing | 16:9 | 480p |
| YouTube or website hero | 16:9 | 1080p |
| TikTok, Reels, Shorts | 9:16 | 720p |
| Cinematic trailer feel | 21:9 | 1080p |
| Final delivery or large screen | 16:9 | 4k |
Reference images, video and audio
This is where the model pulls ahead of simpler generators. The page accepts:
- Up to 9 reference images for character consistency, style and scene layout
- Up to 3 reference videos (15 seconds total) for motion and camera style
- Up to 3 reference audio clips (15 seconds total) for voice and lip sync
Inside your prompt you point to them with tags: [Image1], [Video1], [Audio1]. For example: The woman from [Image1] walks through the market, camera movement copied from [Video1]. Audio references require at least one reference image or video to go with them.

First and last frame control
Upload a first-frame image to animate a still photo. Add a last-frame image and the model works out the motion that connects the two. One rule matters: first and last frames cannot be combined with reference images. Pick one approach per generation.
The seed field makes results reproducible. When you find a clip that is close, keep the seed and change only one phrase in the prompt to see what that phrase does.
Video Length Limits Explained
Duration range
Official provider listings put Seedance 2.0 clips at roughly 4 to 15 seconds. On PicassoIA the duration field defaults to 5 seconds, and you raise it as needed. That is long enough for a single shot with a clear beginning and end, but not for a scene with multiple cuts.

| Duration | Best use |
|---|
| 4 to 5 seconds | Social loops, product reveals, reaction shots |
| 8 to 10 seconds | Establishing shots, short dialogue exchanges |
| 12 to 15 seconds | One continuous action or a slow camera move |
Intelligent duration and getting longer clips
Set duration to -1 and the model picks the length it thinks your prompt needs. This works well for simple actions and poorly when you need a precise timing, so use a fixed number for anything that must hit a beat.
If 15 seconds is not enough, you have two options. The first is to chain clips: take the last frame of one generation, use it as the first frame of the next, and edit the pieces together. The second is to move to a newer model. Seedance 2.5 is reported to double the maximum clip length to about 30 seconds, while Seedance 2.5 Lite targets clips up to 10 seconds. The 2.0 model remains the pick when you need 4K output or lower per-second cost.
💡 Pacing rule: Longer is not better. One clear action per clip beats three rushed ones. If your prompt describes more than two beats, split it into two generations.
What It Costs
Pricing per second
Seedance 2.0 is billed per second of generated video, and the numbers vary by provider, resolution and whether audio is on. These are the figures I found, so treat them as ballpark rather than a quote:
| Source type | Approximate cost |
|---|
| Third-party API, standard model at 480p | About $0.07 per second |
| Third-party API, standard quality | About $0.10 per second |
| Third-party API, Fast mode | About $0.08 per second |
| China domestic pricing | Roughly 0.08 to 0.20 yuan per second |
| Per-generation range on aggregator pages | About $0.60 to $4.05 per clip |

To estimate a project, multiply. At an illustrative $0.10 per second, a 5-second clip is about $0.50, a 10-second clip is about $1.00, and a 15-second clip is about $1.50. Higher resolution pushes that up, and 1080p or 4K clips cost noticeably more than 480p.
💡 Check current rates: Prices change often. Open the Seedance 2.0 model page to see what a generation costs today before you queue a batch.
Ways to spend less
- Draft at 480p. Judge motion and composition cheaply, then re-run the winner at higher resolution with the same seed.
- Use Fast or Mini for tests. Seedance 2.0 Fast and Seedance 2.0 Mini cost less per second.
- Shorten the clip. Generate 5 seconds, not 15, until the idea works.
- Turn off audio if you plan to add your own soundtrack.
- Try the free tier of the newer line. Seedance 2.5 Lite is listed as a free and unlimited generator for clips up to 10 seconds.
How to Use Seedance 2.0 on PicassoIA
Here is the quickest route from idea to finished clip.
- Open the model. Go to the Seedance 2.0 page on PicassoIA.
- Write the prompt. Use the five-part formula: subject, action, camera, light, sound. Put spoken lines in double quotes.
- Choose duration. Start at 5 seconds. Use -1 only if timing does not matter.
- Set resolution. Use 480p or 720p for tests, 1080p or 4k for finals.
- Pick the aspect ratio. 16:9 for desktop, 9:16 for vertical platforms, 21:9 for a widescreen feel.
- Add references if needed. Upload up to 9 images, 3 videos or 3 audio clips, and mention them in the prompt as [Image1], [Video1], [Audio1]. Or use a first-frame image, with an optional last-frame image, but not both approaches together.
- Leave Generate Audio on unless you want a silent clip.
- Generate, review, adjust one thing. Change a single phrase, keep the seed, and run it again.

💡 Workflow tip: For consistent characters across a series, generate a clean portrait first, then reuse it as [Image1] in every clip.
Seedance 2.0 Versus Alternatives
Other video models are worth trying when your priorities differ. All of these are on PicassoIA:
| Model | Where it fits |
|---|
| Seedance 2.0 | Reference-driven clips with native audio and flexible aspect ratios |
| Kling v3 Video | Cinematic motion and strong character movement |
| Veo 3.1 | Polished realism with synchronized sound |
| Sora 2 | Story-driven scenes with synced audio |
| Seedance 2.5 | Longer clips, up to about 30 seconds |

The practical approach is to run the same prompt through two models and keep the better take. Because the prompt formula transfers, you lose nothing by comparing.
Make Your First Clip Today
You now have the prompt formula, the settings that matter, the length range and a way to keep costs in check. The fastest way to find out what works for your own ideas is to run a few short tests. Open Seedance 2.0 on PicassoIA, write a 5-second prompt with one subject and one camera move, and see what comes back. Then change one word and try again.
When you want to try other looks, browse the full lineup at picassoia.com/en/all-models and run the same prompt across a few models. Your best clip is usually the third or fourth attempt, so start now and keep each test short.