As of July 31, 2026, ByteDance has not published an official Seedance 2.5 release date, API endpoint, or price. What is confirmed: ByteDance announced Seedance 2.5 on June 23 at the Volcano Engine FORCE conference, with an "early July" target and a global enterprise beta. What is credibly reported for today: a consumer launch on the Jimeng (即梦) / Dreamina apps, with public API access following roughly a week later (around August 7). Treat the date as a strong signal, not a guarantee — no press release names it. This page updates within hours of launch, and the live price table will be at /models/seedance-2.5.
Below we separate what ByteDance confirmed from the launch-week chatter, then use Seedance 2.0's real rollout and pricing history — every number verified against our dataset — to model what 2.5 is likely to cost.
Confirmed vs Rumor: the state of play
| Claim | Status | Source basis |
|---|---|---|
| Announced June 23, 2026 at Volcano Engine FORCE | Confirmed | ByteDance keynote; Tencent, Huxiu, HK01, SegmentFault (June 23–24) |
| Single-shot native 30-second clips (no stitching) | Confirmed | First-party keynote spec, repeated across all coverage |
| Up to 50 multimodal reference inputs (image/video/audio) | Confirmed | First-party keynote spec |
| Native 4K, 10-bit output | Confirmed | First-party keynote spec |
| Local re-draw / region editing, director-style camera control | Confirmed | First-party keynote demo |
| "Early July" public launch target | Confirmed (as a target) | Volcano Engine release card |
| Launches today (July 31) on Jimeng / Dreamina | Credible | Multiple independent posts, July 30 |
| Public API ~1 week after consumer (≈Aug 7) | Credible | Multiple independent posts + the 2.0 pattern |
| API "went live July 16" on BytePlus | Rumor / likely wrong | Some outlets claimed it; BytePlus ModelArk showed no 2.5 entry as of July 23 |
| ~660 Dreamina credits per 30s 720p clip | Rumor | Single leaked UI screenshot, unverified |
| Specific USD API price | Unknown | No official rate card published |
The most important line in that table is the contradiction. In mid-July, several outlets reported the Seedance 2.5 API had "gone live" on BytePlus. But developer-side checks of the BytePlus ModelArk model list, pricing page, and release notes found no Seedance 2.5 entry as of July 23, and WaveSpeed — a Seedance reseller with every incentive to list it the moment it exists — called 2.5 "previewed but not shipped." The accurate read as of July 31: consumer apps first, API not yet callable, official price not yet posted.
What the signals actually say about timing
The June 23 keynote is the only hard anchor. Everything tighter than "early July" comes from launch-week momentum — real, but not official. Over the seven days ending July 30, "Seedance 2.5" mentions on X ran to roughly 6,000 posts, climbing daily and peaking at ~1,900 on July 30 alone. A single teaser ("5… 4… 3… 2.5 — it's almost here") cleared 900k+ impressions.
Multiple independent accounts — Chinese-language trackers and Western API resellers alike — converged on the same story on July 30: Jimeng consumer launch "tomorrow" (July 31), Volcano Engine and BytePlus API roughly a week out, and a model card and price expected alongside the consumer launch. That is a cross-checked picture, and it is why we are shipping this page today. It is not a press release, so we label it speculation until ByteDance or a first-party dashboard confirms it.
One honest caveat: a launch-day scrum draws affiliate and "coming soon on [platform]" posts, so raw social volume overstates confirmation. We weight first-party keynote facts and multi-source reporting, not hype counts.
What Seedance 2.0's rollout predicts for 2.5
We do not have to guess blind. Seedance 2.0 is fully tracked in our dataset, and its rollout gives a concrete template. Here is what actually happened with 2.0, with every figure pulled from our verified price records.
The launch spread was wide. For the same 720p standard text-to-video clip, 2.0 landed across providers at very different prices: WaveSpeedAI $0.10/s, BytePlus (official) $0.1512/s, Replicate $0.18/s, and fal.ai $0.3034/s. That is a 3.03x gap between the cheapest and most expensive route for an identical output — driven by provider markup and billing scheme, not quality. Expect the same fan-out for 2.5: no single "the price" on day one, but a range.
Resolution scales the bill almost linearly. On Replicate, 2.0 goes from $0.08/s at 480p to $0.18/s at 720p — a 2.25x jump matching the pixel ratio. On fal.ai, 720p is $0.3034/s and 1080p is $0.682/s, again 2.25x. Since 2.5's headline feature is native 4K, the top of its ladder will be materially more expensive per second than any 2.0 720p rate you have seen quoted. A 4K second is roughly 4x the pixels of 1080p.
Which providers moved first. For 2.0, the official BytePlus/Volcano route and the fast-moving aggregators (fal, WaveSpeed, Replicate) carried it early. For 2.5, watch the same set. Our provider probes on July 31 found no live Seedance 2.5 model page yet on WaveSpeed or Replicate, and Dreamina's global page was live in a "coming soon" state — consistent with a consumer-first, API-second rollout.
Modeling a 2.5 price. ByteDance has published nothing, and circulating third-party guesses span a wide $0.022–$0.50 per second. Anchoring to 2.0's real ladder is more defensible: 2.5 will plausibly price at or slightly above the top of the 2.0 per-second range, since 4K/10-bit/30s all demand more compute. At 2.0's verified 720p rates, a 30-second clip would run $3.00 at WaveSpeed's $0.10/s, $4.54 at BytePlus's $0.1512/s, and $9.10 at fal's $0.3034/s — and native 4K pushes the high end well above that. These are 2.0-rate projections onto a 30s clip, not 2.5 quotes.
What to do now (as a developer)
You cannot call a Seedance 2.5 API today, so the highest-value move is cost-planning against known 2.0 rates and keeping your integration flexible.
- Set a real baseline. Run your actual clip volume through /calculator using Seedance 2.0's verified rates. If 2.5 lands near the top of the 2.0 per-second range, that baseline tells you your worst-case bill before you commit.
- Budget for 30-second, 4K economics. The 15s→30s jump doubles output duration, and 4K roughly quadruples per-pixel cost versus 1080p. A workflow that was cheap at 5s/720p can get expensive fast at 30s/4K. Price the specific spec you will ship.
- Keep schemas flexible. No public model ID or rate limits exist yet. Keep a Seedance 2.0 (or alternative) fallback warm so you are not blocked on launch-day capacity.
- Ignore "Seedance 2.5 API" listings that predate the official rate card. Several already circulate; until BytePlus/Volcano posts the model card, treat third-party "it's live" claims as unverified.
- Compare before you migrate. When 2.5 pricing drops, the question is not "is it good" but "is it worth the premium over 2.0 for your use case." Our /cheapest/seedance-2.5 floor tracker answers that the day rates appear.
The promise
The moment Seedance 2.5's price is official, this page and /models/seedance-2.5 update within hours — every rate normalized to USD-per-second so you can compare it against every other model on one axis, the same way we handle the full /models catalog. No estimates left standing once real numbers exist. Bookmark the live tracker and check back on launch day.
Confirmed facts are sourced to ByteDance's June 23 FORCE keynote; the July 31 launch date and API timing are credible multi-source reporting, not official — treated as speculation until ByteDance confirms. Seedance 2.0 rates verified against our dataset July 30, 2026. Rates change frequently — check the live tracker before committing. Some links on this site may be referral links.