Drafting on Seedance 2.0 Today, Locking on Mini at Launch

Mini is in preview. You cannot call it yet. What you can do is shape your production today so that the day the endpoint flips on, you replace one string and the rest of the pipeline carries.

The shape that survives that swap is a four rung ladder. Draft, refine, lock, deliver. Each rung gets its own resolution, duration, and dollar budget. Each rung calls bytedance/seedance-2.0/text-to-video today and will call the Mini endpoint at launch.

Rung one: draft at 480p, 4 seconds

The cheapest rung. You are still picking words.

Parameters: resolution: "480p", duration: 4, aspect_ratio: "16:9", generate_audio: false. Audio off saves wall time and you are not auditioning sound yet.

Cost on Seedance 2.0 today: 4 seconds at $0.3034 per second is $1.21 per draft. Token math gives the same number within rounding. Burn 15 to 20 drafts per shot, total $18 to $24, to find the prompt that holds.

What changes on Mini: per second cost drops. Everything else holds. Preview, subject to change.

Rung two: refine at 720p, 6 seconds

You have a working prompt. Now you stress test it.

Parameters: resolution: "720p", duration: 6, aspect_ratio: "16:9", generate_audio: true. Audio on because dialogue cadence and ambient sound affect how you read the cut.

Cost on Seedance 2.0 today: 6 seconds at $0.3034 is $1.82 per refine. Run 4 to 6 variations on seeds 1 through 6, total $7 to $11 per shot.

What changes on Mini: faster wall clock most likely, lower per second cost. Same prompt grammar should apply. Preview.

Rung three: lock at 720p, 10 seconds

The winning seed and prompt from rung two go here at full target duration.

Parameters: resolution: "720p", duration: 10, aspect_ratio matches delivery, generate_audio: true, seed locked from the rung two winner.

Cost on Seedance 2.0 today: 10 seconds at $0.3034 is $3.03 per lock pass. Render 2 to 3 takes, total $6 to $9 per shot. This is the one you show the client.

Rung four: deliver at 720p, full duration

Same parameters as lock. Different intent. The lock proves the shot. The deliver pass is the file that ships.

Cost: identical to lock, $3.03 for 10 seconds. If the lock is clean, deliver is one pass.

The working call you use across rungs two through four:

01import { fal } from "@fal-ai/client";
02
03// Lock pass on Seedance 2.0 today, swap endpoint when Mini ships.
04const result = await fal.subscribe("bytedance/seedance-2.0/text-to-video", {
05  input: {
06    prompt: "Aerial pull back from a glass office tower at sunset, traffic streaks below, warm key light from the west.",
07    resolution: "720p",
08    duration: 10,
09    aspect_ratio: "16:9",
10    generate_audio: true
11  },
12  logs: true
13});
14
15console.log(result.data.video.url);

What changes when Mini lands

Honestly, less than the marketing will suggest. The most likely deltas, with no specifics until the spec is public:

1. Per second cost drops. The Mini variant of any model family runs cheaper than its parent. By how much: preview, subject to change.
2. Wall clock per call drops. Smaller weights generally mean a faster queue.
3. Prompt grammar carries. The cut beats, named shots, and continuity hints you wrote against Seedance 2.0 should land on Mini with minor tuning.

What does not change: the four rung ladder. Mini does not let you skip the draft rung. Cheaper per call only means you can run more drafts per dollar.

Cost summary per finished shot

On Seedance 2.0 today, a single 10 second shot from draft to deliver lands at $32 to $46 depending on how many drafts and refines you burn. On Mini, the same shot should run lower, with the same shape of ladder underneath. Same rungs, smaller per rung bill.

Lock the ladder now. When Mini opens, you change one endpoint string and keep your prices honest with the client.