How to Use Seedance 2.0 and Seedance 2.5: Settings, Prompts, and Real Credit Costs
How to use Seedance 2.0 on SynthPulse — and why Seedance 2.5, the current flagship, is only about 12% dearer at 1080p. Every resolution and duration setting, what each combination costs in credits, prompt templates for text-to-video and first-frame animation, and how the family compares to Kling.

Most "how to use Seedance 2.0" advice stops at "write a detailed prompt." That skips the three decisions that actually change your output and your bill: which version of the model you run, which direction you run it in, and which resolution and duration you pick.
Start with the version, because it is the one nobody tells you about. Seedance 2.5 is the current flagship of the family, and Seedance 2.0 is the previous generation. Both are live in the Video studio, alongside Seedance 1.5 Pro. 2.0 has not been retired and is genuinely cheaper at the lower resolutions — but if you are shipping 1080p, the gap between them nearly disappears, and picking 2.0 out of habit costs you quality for almost nothing saved.
This guide covers all three, with the real numbers.
Three generations, side by side
Seedance is ByteDance's video family. On SynthPulse all three versions run in both directions — text-to-video from a prompt alone, or image-to-video where your uploaded still becomes the literal first frame of the clip.
At the baseline every other setting multiplies against — 4 seconds, 720p, in credits:
- Seedance 1.5 Pro — 35. The draft tier. Not in the same quality class, and not meant to be.
- Seedance 2.0 — 410. The previous flagship, still available.
- Seedance 2.5 — 630. The current flagship, and the default recommendation in this article.
So at 720p, 2.5 costs 54% more than 2.0. That number is the reason most people default to 2.0 — and it is only true at 720p.
The premium for 2.5 shrinks as the resolution climbs
This is the single most useful fact in this article, and it is not obvious from the model picker. Comparing a 4-second clip at each resolution, in credits:
- 480p — 2.0 is 190, 2.5 is 280. The newer model costs 47% more.
- 720p — 2.0 is 410, 2.5 is 630. 54% more.
- 1080p — 2.0 is 1,020, 2.5 is 1,140. Only 12% more.
The reason is in how the two are priced: 1080p multiplies 2.0's base cost by roughly 2.49, and 2.5's by only about 1.81. The older model is punished harder for pixels than the newer one is.
What follows from that:
- If you are delivering at 1080p, use 2.5. You are paying 120 extra credits on a 1,020-credit clip to run the current flagship instead of last generation's. There is no sensible argument for taking 2.0 at that tier.
- If you are delivering at 720p or 480p, 2.0 is a real saving — 220 credits a clip at 720p, 90 at 480p — and it is a legitimate choice when you are volume-bound rather than quality-bound.
- Either way, do not draft at these prices. Neither of them is the model you discover a prompt on. See the ladder below.
The five settings, in the order they matter
- Version. 2.5 by default; 2.0 when you are staying at 720p or below and counting credits; 1.5 Pro for drafting only.
- Direction. Prompt only, or upload a reference still. For 2.0 and 2.5 the still is sent as the first frame, so the clip begins on exactly that image rather than an interpretation of it. One reference image, not a set.
- Duration — 4, 8, or 12 seconds, for all three versions. Pricing is exactly linear: 8s is 2x and 12s is 3x. There is no bulk discount for a longer take, which means three separate 4-second clips cost the same as one 12-second clip and give you three attempts instead of one.
- Resolution — 480p, 720p, or 1080p, on all three versions. This is where the money is, and, as above, it is also what decides which version to pick.
- Aspect ratio — 16:9, 9:16, or 1:1. And quantity — up to 4 clips per submission, each charged at full price; there is no batch rate.
The credit cost updates on the generate button as you change these, and it is computed by the same code that charges you, so what you see is what you pay.
What every combination costs
Seedance 2.5, in credits:
- 480p / 4s — 280
- 480p / 8s — 560
- 480p / 12s — 840
- 720p / 4s — 630
- 720p / 8s — 1,260
- 720p / 12s — 1,890
- 1080p / 4s — 1,140
- 1080p / 8s — 2,280
- 1080p / 12s — 3,420
Seedance 2.0, in credits:
- 480p / 4s — 190
- 480p / 8s — 380
- 480p / 12s — 570
- 720p / 4s — 410
- 720p / 8s — 820
- 720p / 12s — 1,230
- 1080p / 4s — 1,020
- 1080p / 8s — 2,040
- 1080p / 12s — 3,060
Seedance 1.5 Pro, in credits — the same ladder at roughly a twelfth of 2.0's price and an eighteenth of 2.5's:
- 480p / 4s — 18
- 720p / 4s — 35
- 1080p / 4s — 76
- 1080p / 12s — 226
Read those ladders before you write a single prompt. A 12-second 1080p clip on 2.5 is 3,420 credits — more than a month of the Solo plan, gone on one take. Nobody should be discovering a prompt at that tier.
The workflow that follows: find the shot on 1.5 Pro at 480p for 18 credits, run it five or six times until the motion is right, then re-run the exact same prompt on 2.5 at the resolution and length you actually need. Five drafts plus one 1080p finish is 90 + 1,140 — against 1,140 per attempt if you iterate at the top.
A prompt structure that holds up
Seedance responds to prompts written like a shot description, not like a wish. The structure is the same on all three versions — that is what makes drafting on 1.5 Pro worth anything. Six slots, in this order:
Subject → action → camera → framing → light → grade.
Weak prompt:
a woman walking through a city at night
Same idea, filled in:
A woman in a long charcoal coat walks toward camera along a wet sidewalk. She does not look up. Camera tracks backward at walking pace, staying level with her chest. Medium shot, 35mm, shallow depth of field. Lit only by shopfront neon on the left and a streetlight behind her, so her face falls in and out of light as she moves. Cool blue shadows, warm highlights, slight film grain.
Three more worked examples you can adapt:
Product
A matte black ceramic mug sits on a pale oak table. Steam rises steadily from it. Camera orbits slowly 30 degrees to the right, at table height. Soft window light from the left, one shadow only. Clean, neutral color, nothing else in frame.
Character moment
An old man at a kitchen table looks down at a photograph in his hands, then slowly raises his head toward the window. Camera holds completely still, medium close-up, eye level, 50mm. Late afternoon light through a half-closed blind lays stripes across his face. Muted palette, no camera movement at all.
Landscape
Drone rises vertically from a pine forest floor through the canopy and out above the treetops, revealing a valley in low fog. Continuous upward movement, no cuts, no tilt. Overcast light, flat contrast, desaturated greens.
Three patterns run through all of these. Every prompt names the camera behaviour explicitly, because a model left to guess will usually invent a drift you did not want. Every prompt describes one continuous action, because asking for a cut inside a 4-second clip gets you a smear where the cut should be. And every prompt says what stays fixed, not only what moves.
Image-to-video: animate a still you already control
Text-to-video means re-rolling the composition every attempt. First-frame animation means locking the composition once and only iterating on the motion — usually cheaper and always more predictable. It costs the same as text-to-video on every Seedance version; nobody charges extra for starting from a frame.
The two-step version:
- Generate the still in the Image studio. Nano Banana Pro (45 credits) and Seedream 5.0 Pro (18 credits at 1K) are both good here; GPT Image 2 (15 credits) is the one to reach for when the frame has readable text in it.
- Bring it into the Video studio, attach it as the reference, and write a prompt that describes only the motion.
That last part is where most people lose the frame. Do not re-describe the subject — the model can already see it, and repeating the description invites it to redraw what you carefully composed. Compare:
Bad: A red vintage car on a coastal road at sunset, driving fast, cinematic.
Good: The car pulls away from camera down the road. Camera stays locked off. Only the car, the wheels, and the grass at the roadside move. Everything else holds still.
One limit to plan around: SynthPulse sends a single reference image as the first frame, on 2.0 and 2.5 alike. Multi-image references are not part of this path, so the composition you want has to exist in one still.
Your clips come back silent
SynthPulse submits every Seedance version with audio generation explicitly switched off, and the credit prices quoted above are the no-audio tier. Plan on adding sound afterwards — an editor, or a track from the Song studio, which shares the same credit balance.
The one exception on the platform is Veo 3.1 Fast, the only video model here that SynthPulse does not silence on submit. If a shot lives or dies on dialogue, that is the model to use for it.
Seedance vs Kling
This is the comparison people actually want, so here it is at 720p, in credits:
- Kling 2.5 Turbo Pro — 105 for 5 seconds, 210 for 10 seconds
- Kling 2.6 — 138 for 5 seconds, 275 for 10 seconds
- Kling 3.0 Pro — 140 for 4 seconds, 280 for 8 seconds
- Seedance 2.0 — 410 for 4 seconds, 820 for 8 seconds
- Seedance 2.5 — 630 for 4 seconds, 1,260 for 8 seconds
At 1080p the spread widens sharply: Kling 3.0 Pro is 180 credits for 4 seconds, against 1,020 for Seedance 2.0 and 1,140 for Seedance 2.5. That is 6.3x for the flagship.
So the honest framing is not "which model is better." It is: the Kling family gets you 1080p and 10-second takes for a fraction of the price, and Seedance has to earn a 6x premium on the specific shot in front of you. Sometimes it does — complex physical motion and multi-element scenes are where it tends to pull ahead. Often it does not. The way to find out costs 18 credits a try: run the same prompt on Seedance 1.5 Pro and Kling 3.0 Pro at their cheapest tiers and look.
One structural difference worth knowing: Kling 2.6 and Kling 2.5 Turbo Pro accept only 5- or 10-second durations, and only 720p. Every Seedance version accepts 4, 8, or 12 seconds at any of the three resolutions. If your edit needs a 12-second unbroken take, Seedance is the family that can make it.
Before you hit generate
- Shipping at 1080p? Then it is 2.5, not 2.0 — 12% more for the current flagship.
- Staying at 720p and counting credits? 2.0 saves you 220 a clip. That is the only place the choice is close.
- Drafting? Neither. 1.5 Pro at 480p, 18 credits a try.
- Does the prompt name the camera, or did you leave it to chance?
- One continuous action, no cuts?
- Have you said what stays still?
- If you have a composition in mind already, are you animating a still instead of re-rolling it?
More on the family: Seedance on SynthPulse. The cheaper long-take alternative: Kling. Picking between the top-end models: the shortlist by use case, or the six questions that narrow it down. Or just go make something — open the Video studio.
