Seedance 2.5 Flash on WaveSpeedAI: Specs, Access, and When to Use It

Seedance 2.5 Flash is a new, faster tier of ByteDance’s Seedance 2.5 video model. ByteDance’s Volcano Engine announced it on November 12, 2026, as the model closest in quality to Seedance 2.5, with generation speed doubled and a lower price.

Seedance 2.5 Flash is available on WaveSpeedAI. Start from the Seedance 2.5 Flash page, or go straight to text-to-video, image-to-video, reference-to-video or video-edit. This post covers what ByteDance announced, how to use Flash on WaveSpeedAI, and how to choose between Flash and Seedance 2.5.

What Did ByteDance Announce?

According to ByteDance’s announcement as reported by IT之家 and PANews:

• Speed: generation is twice as fast as Seedance 2.5.
• Price: lower than Seedance 2.5, by about a third at 720p text-to-video and image-to-video.
• Quality: overall better than the previous flagship, Seedance 2.0, with strong instruction following, physical motion and audio, and support for video editing.
• Output: up to 30 seconds per video, at 480p or 720p.
• Inputs: up to 10 images, 5 videos and 5 audio clips per request.
• Positioning: built for high-volume work such as testing ad creative and producing regular footage, while Seedance 2.5 stays the choice for the highest audio-visual quality, Draft mode, 1080p and 4K.

BytePlus’s Lumina preview page describes Flash as a speed-focused addition to the Seedance family, with international access, API and pricing still to be confirmed.

Announced vs. Still Open at a Glance

Question	Status
Is Flash announced?	Yes, November 12, 2026
What is it for?	Faster, lower-cost generation for high-volume work
Speed	Announced as 2x Seedance 2.5; no independent measurements yet
Clip length and resolution	Up to 30 seconds; 480p and 720p
Inputs	Up to 10 images, 5 videos, 5 audio clips; video editing supported
ByteDance API	Scheduled for Volcano Ark on November 15
Available on WaveSpeedAI	Yes: text-to-video, image-to-video, reference-to-video and video-edit

What Should You Not Assume?

Announcement numbers are a starting point, not a test result:

• That ByteDance’s “2x faster” holds for your prompts. Speed depends on duration, resolution, references and queue time. Measure on your own shots.
• That Flash matches Seedance 2.5 on every feature. Flash tops out at 720p, so 1080p and 4K work still belongs to Seedance 2.5. Check duration, references, audio and editing one by one when you test it.
• That a lower price means a lower project cost. What matters is the cost of an accepted result, including the takes you throw away.
• That “Flash” means the same thing as “Fast”, “Mini” or “Turbo”. These are different tiers with different limits. Compare them by their published settings, not by their names.

How Do You Use Flash on WaveSpeedAI?

Flash has four endpoints, all listed on the Seedance 2.5 Flash page. Text-to-video, image-to-video and reference-to-video generate 4 to 30 second clips at 480p or 720p (default 720p) with synchronized audio on by default; video-edit follows the length of the input video:

Endpoint	What it does
text-to-video	Prompt in, clip out. Optional reference images, videos and audio.
image-to-video	Animates a start image, with an optional last frame.
reference-to-video	Up to 10 reference images, 5 reference videos and 5 reference audios, cited in the prompt as @Image 1, @Video 1, @Audio 1.
video-edit	Changes an existing video from a prompt while keeping its subject, composition and motion.

Aspect ratios are 16:9, 9:16, 4:3, 3:4, 1:1 and 21:9 for text-to-video and reference-to-video. Billing is per second of generated video and depends on resolution; see each model page for current prices.

curl -X POST "https://api.wavespeed.ai/api/v3/bytedance/seedance-2.5-flash/text-to-video" \
  -H "Authorization: Bearer $WAVESPEED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "A ginger cat pounces on a ball of yarn on a sunlit wooden floor. Low eye-level shot, warm afternoon light, soft ambient room tone.", "resolution": "720p", "duration": 5}'

curl "https://api.wavespeed.ai/api/v3/predictions/<id>/result" \
  -H "Authorization: Bearer $WAVESPEED_API_KEY"

What Else Can You Use on WaveSpeedAI?

These other Seedance models are live too, all with native audio generated together with the picture:

Model	Good for	Output
Seedance 2.5 Flash	Fast iteration and high-volume clips	4 to 30 seconds, 480p or 720p
Seedance 2.5 Turbo	A faster, more affordable Seedance 2.5 tier for everyday iteration	4 to 30 seconds, 720p or 1080p
Seedance 2.5	Final shots and the full feature set	4 to 30 seconds, 480p to 4K
Seedance 2.0 Fast and Seedance 2.0 Mini	Lower-cost drafts on the previous generation	See each model page

Seedance 2.5 Turbo is another faster tier, with 720p and 1080p output. It takes the same inputs as Seedance 2.5 text-to-video: a prompt, up to 30 reference images, up to 10 reference videos and up to 10 reference audios, six aspect ratios from 9:16 to 21:9, and audio on by default. There is also an image-to-video Turbo model that starts from a first frame and optionally ends on a last frame.

Billing is per second of generated video and depends on resolution, with no subscription. Current prices are on each model page and on the Seedance 2.5 API page. If you prefer the browser, the AI Video Generator runs the same models without code.

A Draft-Then-Final Workflow

Flash is aimed at a simple way of working: iterate cheaply, then render the keeper once.

1. Draft on Flash. Use Seedance 2.5 Flash at 480p or 720p with a short duration to settle the idea: framing, action, camera move and sound.
2. Change one thing per take. Compare a static camera against a slow push-in, or two lighting setups. One variable per take makes the choice obvious.
3. Lock the prompt and references. Once a take works, stop editing the prompt.
4. Render the final once. For 1080p or 4K delivery, send the same prompt and references to Seedance 2.5. If 720p is enough, the Flash take is already your final.

The request is the same for both steps. Only the model ID and the resolution change:

MODEL="bytedance/seedance-2.5-flash/text-to-video"   # 1080p/4K final: bytedance/seedance-2.5/text-to-video

curl -X POST "https://api.wavespeed.ai/api/v3/$MODEL" \
  -H "Authorization: Bearer $WAVESPEED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "A ceramic mug on a pale wooden table by a window. A hand places a spoon beside it. Slow push-in, soft morning light from the left.", "aspect_ratio": "16:9", "resolution": "720p", "duration": 5}'

curl "https://api.wavespeed.ai/api/v3/predictions/<id>/result" \
  -H "Authorization: Bearer $WAVESPEED_API_KEY"

For prompt structure, see the Seedance 2.5 prompt guide.

How to Test Flash Against Your Current Model

• Keep the model ID configurable. If the model is a setting and not a hardcoded string, switching to Flash is a one-line change.
• Save a small test set. Pick 10 to 20 prompts that look like your real work: a product shot, a dialogue scene, a fast action shot, a scene with several characters. Keep the reference files with them.
• Record a baseline. For each prompt, note the time from request to result, the cost, and whether the take was usable on your current model. Without those numbers you cannot tell what Flash improves.
• Write down your acceptance checks. Subject stays consistent, the action completes, the framing is stable, the last frame is usable. Score every model against the same list.

How to Judge Flash

Run the same test set on Flash and on the model you use now, with matching duration, resolution and references wherever both support them. Then compare three numbers:

• Time to a usable clip, including queue time and retries, not only generation time.
• Usable-output rate: clips that pass every check divided by clips generated.
• Cost per accepted clip: total spend divided by the clips you would actually ship.

A model that is twice as fast but needs twice as many takes has not saved anything. Use several representative shots, not one favorable example.

FAQ

Is Seedance 2.5 Flash released?

It was announced by ByteDance on November 12, 2026, and is rolling out on ByteDance’s own apps and partner platforms. Its API is scheduled to open on Volcano Ark on November 15.

Is Seedance 2.5 Flash available on WaveSpeedAI?

Yes. Seedance 2.5 Flash text-to-video, image-to-video, reference-to-video and video-edit are available, alongside Seedance 2.5, Seedance 2.5 Turbo, Seedance 2.0, Seedance 2.0 Fast and Seedance 2.0 Mini.

How is Flash different from Seedance 2.5?

ByteDance positions Flash as twice as fast and lower in price, with 480p and 720p output, while Seedance 2.5 keeps 1080p, 4K and Draft mode for the highest-quality work.

Will Seedance 2.5 Flash be cheaper than Seedance 2.5?

ByteDance says Flash is priced lower, and on WaveSpeedAI Flash is priced below Seedance 2.5 at the same resolution. Check each model page for current prices, compare at the same duration and resolution, and count the takes you discard.

Is Seedance 2.5 Turbo the same as Seedance 2.5 Flash?

No. Seedance 2.5 Turbo is a faster, more affordable Seedance 2.5 tier with 720p and 1080p output. Flash is a separate model with 480p and 720p output.

What is the fastest way to use Seedance today?

Draft with Seedance 2.5 Turbo at 720p, then render the final take with Seedance 2.5 at 1080p or 4K.

Details about Seedance 2.5 Flash come from ByteDance’s announcement as reported by news outlets and may change. Recheck the official documentation and the model pages before planning production work around it.

---

Related posts:

• Seedance 3.0 rumors: what to expect
• Seedance 2.5 prompt guide