Seedance 2.5 Economy API

Seedance 2.5 Economy is the efficiency- and cost-focused video generation model of ByteDance's Seedance 2.5 series. Seedance 2.5 launched on July 31, 2026 on Jimeng and Dreamina (with the Volcano Engine API opening on August 7, 2026), known for native single-pass generation up to 30 seconds; Economy carries that capability with a variable 4–30 second duration (default 5, or -1 to let the model choose automatically), supporting text prompts plus image, video, and audio references, cited in the prompt via @Image1, @Video1, and @Audio1 markers. This "Economy" version of the model does not support real human faces; please try the official version if you need to generate videos featuring real human faces.

The model is offered through the iCreat platform as an API using a two-step asynchronous task flow: submit a task to obtain a task_id, then poll the result endpoint until a terminal status. Resolution supports 480p / 720p / 1080p with seven aspect ratios (including adaptive); billing follows "unit price × (output video duration + reference video duration)" — reference video duration is billable, while image and audio references are not.

Model Positioning

Seedance 2.5 Economy targets "flexible-duration, cost-first video production." Unlike the fixed-30-second Economy 30s, this tier lets duration range freely from 4 to 30 seconds — suited to short clips, storyboard segments, and dynamic assets of varying lengths; and compared with the flagship Seedance 2.5, it trades a more focused capability set for a better per-second price. Batch production where the cost-to-duration ratio must be controlled precisely — e.g., generating segment by segment to a script, then stitching — is its most typical use.

Core Capabilities

Flexible 4–30 Second Duration

Specify duration as needed: any integer from 4 to 30 seconds (default 5), or pass -1 and let the model decide based on the content — one interface covering both long and short pieces.

Omni-Modal Reference Generation

Supports image, video, and audio references: anchor subject appearance with a reference image, transfer camera movement with a reference video, and align rhythm with reference audio — cited in the prompt via @ImageN, @VideoN, @AudioN markers.

Three Resolution Tiers

480p / 720p / 1080p selected as needed, balancing cost and clarity by stage: iterate drafts at 480p, deliver finals at 1080p.

Seven Aspect Ratios Including Adaptive

ratio supports six concrete values — 16:9, 4:3, 1:1, 3:4, 9:16, 21:9 — plus adaptive, where the model automatically selects a suitable ratio based on the input content.

Transparent Billing

Total price = unit price × (output duration + reference video duration); image and audio references are excluded from duration, so costs are precisely predictable.

Pricing

Total Cost = Unit Price × (Output Video Duration + Reference Video Duration). With no reference video, only the output duration is billed; reference video duration is included in the billable total when provided; image and audio references are not included. The costUSD field in the response returns the actual cost once the task succeeds.

Application Scenarios

• Script-segmented generation: produce segment by segment to storyboard seconds, then stitch into long videos
• Short-form and feed content: batch production of 4–15 second vertical clips
• Multi-asset mixed generation: creative content driven by subject image + camera video + rhythm audio
• Two-stage cost iteration: validate drafts at 480p, output finals at 1080p
• Reference-driven re-creation: deriving new videos from existing assets

Model Comparison

Same-Series Comparison

Cross-Model Comparison

Same-series data comes from ByteDance's official releases; this model's specifications follow the iCreat channel documentation.

Why Choose Seedance 2.5 Economy?

• Duration specified freely from 4 to 30 seconds, or -1 for the model to decide — one interface for long and short content
• Omni-modal references with @ markers: subject, camera movement, and rhythm anchored by image, video, and audio
• Three resolution tiers on demand, with cost controlled by stage between drafts and finals
• Seven aspect ratios including adaptive, covering delivery sizes across platforms
• Transparent billing: unit price × (output + reference video duration); image and audio references free of duration charges

Specifications

Architecture

Seedance 2.5 is ByteDance's next-generation video generation model released on July 31, 2026, extending single-pass duration from Seedance 2.0's ~15 seconds to a native maximum of 30 seconds. The series supports up to 50 multimodal references (30 images + 10 videos + 10 audio clips), long prompts with second-level timestamps, targeted editing of specified segments, and videos generated with synchronized audio. Economy is the series' variable-duration tier for efficiency and cost: it retains omni-modal referencing, lets duration range from 4 to 30 seconds or be decided by the model, and serves volume production with three resolution tiers and transparent billing.

Notes

• Submit and query must be chained with the same task_id
• While processing, result is []; keep polling, with a suggested interval of 2–5 seconds
• FAILED is terminal; check request parameters and reference media before resubmitting
• duration accepts integers 4–30, or -1 for the model to choose; it defaults to 5 when omitted
• Reference video duration is billable: with a reference video, total price = unit price × (output duration + reference video duration)
• Image and audio references are not included in the video duration
• The order of reference arrays must match the @ indexes used in the prompt
• When ratio is adaptive, the model selects the aspect ratio automatically based on input content
• All reference media URLs must be publicly accessible
• costUSD is returned only when the task succeeds; failed tasks do not produce a cost field

FAQ

How do I get started with the Seedance 2.5 Economy API?​

Register on the iCreat platform and obtain an API Key from the console, send a generation request to the submit endpoint, then poll the result endpoint with the returned task_id. All requests are authenticated with the Authorization: Bearer header. Keep your API Key safe and never expose it in client code or public repositories.

How is the cost calculated?​

Total Cost = Unit Price × (Output Video Duration + Reference Video Duration). For example, a 720p 10-second video with no reference video costs 0.303 × 10 = $3.03; with a 5-second reference video, it becomes 0.303 × (10 + 5) = $4.545. Image and audio references are not included in the duration. The costUSD field in the response gives the actual cost once the task succeeds.

How do I set the duration?​

Pass an integer from 4 to 30 in duration for an exact length (default 5), or pass -1 and let the model choose a suitable duration from the prompt content. When stitching long videos segment by segment, explicitly specifying seconds per segment keeps duration predictable.

How do I use reference media?​

Images, videos, and audio are passed via the images, videos, and audios arrays respectively, and cited in the prompt with @Image1, @Video1, @Audio1 markers (indexes start at 1, matching array order). For example: "Use the subject from @Image1, the camera movement from @Video1, and the rhythm from @Audio1 to generate a video."

How do I choose between this and Economy 30s?​

Two routes: if content needs variable 4–30 second durations, adaptive aspect ratios, or the 1080p tier — choose this model; if content is fixed at 30-second complete short videos and minimum per-second cost is the goal — choose Economy 30s (its 720p per-second cost is significantly lower).

What if the task fails?​

FAILED is a terminal status. Check in order: whether the prompt is empty, whether reference media URLs are publicly accessible, whether @ indexes match the array order, whether resolution is one of the three allowed values, and whether duration is an integer 4–30 or -1. Fix any issue and resubmit. The cost field is returned only when the task succeeds.