Seedance 2.5 Flash: ByteDance's faster, cheaper Seedance tier is live in apps now, with the API on Nov 15
ByteDance has added a Flash tier under Seedance 2.5. BytePlus says it gets close to flagship 2.5 quality at nearly twice the speed and about 34% lower cost per second at 720p. It tops out at 720p, keeps 30-second clips and multi-reference input, and is already live in Higgsfield, CapCut and Lovart. The ModelArk API opened on Nov 15; API notes below.
What launched. On Nov 11 at 11:09 PM MST, BytePlus announced Dreamina Seedance 2.5 Flash, a faster and cheaper tier of the Seedance 2.5 video model. The vendor's pitch: results 'close to flagship Seedance 2.5', nearly twice the generation speed, and USD 0.152 per second for 720p text-to-video and image-to-video, which it says is about 34% below Seedance 2.5. The API goes live on BytePlus ModelArk on Sunday, Nov 15 at 14:00 UTC+8 (Saturday, Nov 14 at 11 PM MST) under the model ID dreamina-seedance-2-5-flash-261115.
What it can do, per BytePlus's docs. Clips up to 30 seconds. 480p and 720p only, with no 1080p or 4K, so the flagship is still the one for final masters. Text-to-video, image-to-video and multi-reference input with up to 10 images, 5 video clips and 5 audio clips per request. The flagship takes 30 images, 10 videos and 10 audio clips. ModelArk lists a 5-second 16:9 clip at about $0.34 at 480p and $0.76 at 720p, against about $1.16 for flagship 2.5 at 720p. Video input costs more, since tokens count input plus output duration.
Where you can use it today. App partners switched it on before the API. Higgsfield launched it in early access on Nov 12 (about 335 likes and 167 reposts by mid-morning), says it isn't available in the US yet, and paired it with a sale. CapCut's AI Lab advertises it at $0.01 per second, 50% below Seedance 2.5. That's CapCut's US app pricing and it varies by plan and market. Lovart and Renoise have it too, and WaveSpeed says it's coming soon. App prices are each platform's own, so check the per-clip cost where you generate.
Early reactions. The first side-by-side we saw came from @DDJCXX on Jimeng, ByteDance's Chinese app. It used the same image and prompt for 5 seconds on each model. He found Flash clearly faster but rated flagship 2.5 better, pointing to more natural foam and swirl in a glass of orange juice. That's one comparison by one person, not a benchmark, and we haven't run our own.
How it fits a pipeline. Flash looks built for the drafting stage: blocking shots, testing prompts, trying ad concepts in bulk and checking reference setups at 480p/720p before paying for the flagship (or another model) on the takes you keep. It's the same draft-then-master split Google pitches with Omni 1.2 Flash and Kling with 4.0 Flash. The 720p ceiling means you'll still need an upscaler or a flagship re-render for delivery.
API notes (updated Nov 16). BytePlus's API reference now says the Seedance 2.5 series, including Flash, is fully available for API calls and online trials on ModelArk. Calls use the same async video-task endpoint as other Seedance models (POST /api/v3/contents/generations/tasks, then poll the task) with model set to dreamina-seedance-2-5-flash-261115 and API-key auth. Resolution defaults to 720p, with 480p as the only other option. duration accepts 4 to 30 seconds, or -1 to let the model choose, and -1 is the default for the 2.5 series. An omni_reference_task_type hint (auto, reference, edit or extend) controls how reference videos are treated. Edit jobs need a 4-30 second reference video, an adaptive ratio and duration -1. Unlike Seedance 2.0, the 2.5 series can work from audio-only references. One gap for drafting: the docs say Draft mode (a quick 480p preview you can later promote to a final render) works only on flagship Seedance 2.5, not on Flash. For Flash, just render at 480p and re-render the keepers. These come from BytePlus's docs. We haven't made test calls.
Signals: official launch by BytePlus plus same-day rollouts by tracked official @higgsfield and other platforms. No tracked curator pick or GitHub signal yet, so it's not HOT by our rule.
Related tools on ReelStack
Skills & repos in our directory
- Comfy Previs · Repo, Generation
- seriesmith · ComfyUI, Generation
More from the Lab
- Weekend release run: HyperFrames 0.9.43 to 0.9.58, Remotion 4.0.544 adds OpenTimeline export, and Seedance 2.5 Flash lands in ComfyUI
- Topview T1: a new video model priced against Seedance 2.5, and the first independent test measured 26% cheaper, not 62%
- Kling Image 4.0 launches with native 8K and image sets, Kaggle opens a Code2Video Arena, and HyperFrames reaches 0.9.63
- A two-layer launch video: code motion over a Seedance plate
Topics: Higgsfield · Seedance