Seedance 2.5 Flash has four routes: Jimeng and partner platforms (consumer), the Volcano Ark API (official, domestic), the BytePlus international API, and WaveSpeedAI (third-party aggregator). There are no weights and no local running — every call goes through an API or an app. This article covers how each route works, what to watch when calling it, and the Flash-vs-flagship trade-off.
| Route | Entry point | Live since | For whom |
|---|---|---|---|
| Jimeng and partner platforms | Jimeng, LibTV, Updream, Creao AI, TapNow | Nov 12, 2026 | Users who don't want to write code; per-call use |
| Volcano Ark API | Model ID doubao-seedance-2.5-flash | Nov 15, 2026, 14:00 | Domestic developers; billed per second |
| BytePlus international API | Model ID dreamina-seedance-2.5-flash-261115 | Nov 15, 2026 | International users; token-priced |
| WaveSpeedAI | Third-party aggregator | Listed Nov 13, 2026 | Users who don't want to integrate the official APIs directly |
The model ID is doubao-seedance-2.5-flash, live since Nov 15, 2026 at 14:00. Text-to-video and image-to-video cost the same: 720p ≈ ¥0.99/s, 480p ≈ ¥0.44/s. Confirm four things before calling: resolution is 480p/720p only — a 1080p request will fail or be rejected; the maximum is 30s per generation, so split anything longer; the reference cap is 20 per request (≤10 images, ≤5 videos, ≤5 audio clips), so count references in multi-asset prompts first; and there is no Draft mode — for a cheap-draft-then-final workflow, use the 480p tier (¥0.44/s) as the draft and render the final at 720p, which is Flash's own low-cost workflow. Built-in synchronised audio is kept, so finished clips need no separate soundtrack; video editing is supported for quick variants without full regeneration.
International users go through BytePlus ModelArk, model ID dreamina-seedance-2.5-flash-261115, live the same day as the domestic API, Nov 15, 2026. Billing differs from domestic: token-priced at $7.0 per 1M tokens at 480p/720p (flagship $10.70); a 5s 720p clip ≈ $0.76, about $0.15/s — only one conversion, with Atlas Cloud's reference-to-video row listing $0.167/s. When citing a USD figure, state the conversion basis — never a single number alone. WaveSpeedAI lists Flash as available, for users who don't want to integrate BytePlus directly, billed on its own platform pricing and quotas.
Three things Flash is for: drafts and batches — 480p at ¥0.44/s, cheapest for generating many clips and picking one; regular finished clips where 720p is enough — 10s ≈ ¥9.9, 30s ≈ ¥29.7, ~34% below the flagship; and long batch clips — the 30s maximum is kept, so long batch clips lose nothing. Three things it isn't for: high-resolution finals — the ceiling is 720p, so 1080p/4K needs the flagship Seedance 2.5; the Draft workflow — Flash has no Draft mode, so the one-click draft → approve → final flow needs the flagship; and complex multi-subject prompts — the 20-reference cap means anything above it needs the flagship (50). In one line: Flash is the draft-and-batch tier, 2.5 is the finals tier.
Domestic Ark pricing: a 10s 720p clip ≈ ¥9.9, 30s ≈ ¥29.7; 10s 480p ≈ ¥4.4, 30s ≈ ¥13.2. Ten 10s 720p clips ≈ ¥99; ten 30s 720p clips ≈ ¥297. International BytePlus pricing: a 5s 720p clip ≈ $0.76, 10s ≈ $1.5 (at the ~$0.15/s conversion; see the pricing guide for the basis). There is no free tier — budget at full price; Seedance 2.0 is the only Seedance model with a confirmed free tier.
Domestic Volcano Ark: doubao-seedance-2.5-flash; international BytePlus ModelArk: dreamina-seedance-2.5-flash-261115.
Jimeng has offered the Flash tier since Nov 12 — open Jimeng and it is selectable. LibTV, Updream, Creao AI and TapNow went live the same day.
No. Flash is a closed model with no public weights — API and app access only.
Use the 480p tier (¥0.44/s) as the cheap draft, then render the final at 720p. That is Flash's own low-cost workflow — without the flagship Draft mode's one-click flow.
Reference: Volcano Ark (model ID doubao-seedance-2.5-flash, Nov 15, 2026) · BytePlus ModelArk (model ID dreamina-seedance-2.5-flash-261115, Nov 15, 2026) · WaveSpeedAI (Nov 13, 2026) · Startup Fortune (Nov 15, 2026) · Atlas Cloud (ByteDance provider page) (rewritten and organized by ModelBrief; model IDs follow the official announcement)