
Seedance 2.5 Flash API
seedance-2.5-flashSeedance 2.5 Flash is the faster tier of ByteDance's Seedance 2.5 video model: it turns a prompt, a first frame or up to 20 reference files into a 4–30 second clip at 480p or 720p with sound generated alongside the picture. ByteDance says Flash generates about twice as fast as Seedance 2.5 and describes it as the model whose results come closest to Seedance 2.5. On APIMODELS it is the model id seedance-2.5-flash on POST /v1/video/generations, from $0.091 per second at 480p, and failed tasks are not charged. Each clip is any whole length from 4 to 30 seconds, in 1:1, 4:3, 3:4, 16:9, 9:16, 21:9 or adaptive, delivered as mp4 or mov; sound is on by default and generate_audio: false turns it off. Text-to-video needs only a prompt. Image-to-video starts from first_frame_url, with last_frame_url as an optional end frame. Reference-to-video takes up to 10 images in reference_image_urls, up to 5 videos in reference_video_urls (each 2 to 30 seconds, 30 seconds in total) and up to 5 audio clips in reference_audio_urls. Frames and references cannot be combined in one request. ByteDance's documentation says the Seedance 2.5 series does not accept reference images or videos showing real human faces as direct uploads. On APIMODELS, real-person video goes through the same asset library as Seedance 2.5: register the portrait once, then put its asset:// reference in first_frame_url to animate that person, or in reference_image_urls to borrow the look; a raw portrait URL is rejected when the task is created and is not charged. Billing is per second of output: $0.091 at 480p and $0.195 at 720p, so a 5-second clip costs $0.455 at 480p and $0.975 at 720p, and a 10-second clip $0.91 and $1.95. When a request includes reference video the rate is $0.056 at 480p and $0.119 at 720p per billable second, where billable seconds are the total length of the uploaded reference videos, each clip rounded up to a whole second, plus the generated length: a 10-second 720p clip with a 5-second reference video is 15 billable seconds, $1.785. Reference images and audio add nothing. For comparison, Seedance 2.5 on APIMODELS is $0.120 per second at 480p and $0.270 at 720p ($0.072 and $0.162 with a reference video), so Flash runs at a lower rate per second, about 24% less at 480p and 28% less at 720p; a 10-second 720p clip is $1.95 on Flash and $2.70 on Seedance 2.5. Seedance 2.5 takes up to 30 images, 10 videos and 10 audio clips per job, against 10, 5 and 5 on Flash. Flash also takes automatic duration and the edit and extend task types. With duration: -1 the model picks a length from 4 to 30 seconds; the request reserves 30 seconds at creation ($5.85 at 720p, $2.73 at 480p) and is charged only for the delivered length, so a 720p clip that comes back 12 seconds long costs 12 × $0.195 = $2.34. task_type edit takes the source clip in reference_video_urls, keeps its length (duration -1 or omitted) and is billed at the reference-video rate on source plus output seconds: editing an 8-second clip at 720p is (8 + 8) × $0.119 = $1.904. task_type extend works the same way with a duration of 4 to 30 or -1: extending an 8-second clip by 6 seconds at 480p is (8 + 6) × $0.056 = $0.784. Edits and extensions need aspect_ratio adaptive or left out. These three may occasionally return a 400 that is not charged; seedance-2.5 offers them too. There is no 1080p, no seed parameter and no web search tool. Typical uses: short social clips in batches, drafts of a shot before the final render, quick edits and extensions of existing clips, animating a product or a character from a first frame, a consenting real person from the asset library, and new clips that follow the camera movement of a reference video or the music of a reference audio clip. Not every request succeeds: content review or upstream queuing can fail a task, and failed tasks are not charged. In single tests on 2026-11-17, a 480p 4-second text-to-video clip finished in about 70 seconds, and a 480p 4-second clip from one reference portrait in the asset library in about 61 seconds; those are single tests, not a guarantee. The API is asynchronous: the create call returns a task id, and you poll GET /v1/video/generations?task_id= until the state is completed, or pass callback_url.
Generate Video
Loading library…
mp4 兼容性最好;mov 是 H.264 + yuv444p + PCM,色彩还原更准,适合调色 / 抠像 / 合成,但部分播放器打不开。
Render synced audio alongside the video
Result
Generated video will appear here
Provide URLs and click Generate
4-30 Seconds in One Pass
Any whole length from 4 to 30 seconds from a single request, at 480p or 720p, in 1:1, 4:3, 3:4, 16:9, 9:16, 21:9 or adaptive, as mp4 or mov
Native Synced Audio
Sound is generated together with the picture by default. Pass generate_audio: false when you want a silent clip
Text, Frames or up to 20 References
A prompt alone, a first frame with an optional last frame, or up to 10 reference images, 5 videos (2-30 seconds each, 30 seconds in total) and 5 audio clips. Frames and references cannot be combined in one request
Real Person via the Asset Library
ByteDance does not take raw real-face uploads for the 2.5 series. Register the portrait once and pass its asset:// reference, in first_frame_url to animate that person or in reference_image_urls to borrow the look. Consented subjects only
From $0.091 per Second
480p $0.091/s and 720p $0.195/s, against $0.120/s and $0.270/s for Seedance 2.5 on this site. Reference images and audio add nothing
Reference Video Billing
$0.056 (480p) or $0.119 (720p) per billable second: the total length of your reference videos, each clip rounded up to a whole second, plus the generated length
Edit, Extend and Automatic Length
task_type edit or extend on a source clip, billed on source + output seconds at the reference-video rate; duration -1 reserves 30 s and charges the delivered length
Limits, Stated Plainly
No 1080p, no seed and no web search. Content review or upstream queuing can fail a request, and edit, extend or -1 requests may occasionally return a 400; failed tasks are not charged
Seedance 2.5 Flash API: Seedance 2.5 video with sound, from $0.091 per second
Seedance 2.5 Flash is the faster tier of ByteDance's Seedance 2.5 video model: it turns a prompt, a first frame or up to 20 reference files into a 4–30 second clip at 480p or 720p with sound generated alongside the picture.
ByteDance says Flash generates about twice as fast as Seedance 2.5 and calls it the model that comes closest to Seedance 2.5 in output quality, ahead of Seedance 2.0.
On APIMODELS the Seedance 2.5 Flash API is the model id seedance-2.5-flash on POST /v1/video/generations, from $0.091 per second at 480p, and a Seedance 2.5 Flash task that fails is not charged.
- Model ID
seedance-2.5-flash- Duration
- 4–30 s per clip
- Resolution
- 480p · 720p
- References
- 10 images · 5 videos · 5 audio clips
- Price
- $0.091/s (480p) · $0.195/s (720p)
Seedance 2.5 Flash at a glance
Every limit of one Seedance 2.5 Flash request on APIMODELS, in one table.
| Item | Seedance 2.5 Flash on APIMODELS |
|---|---|
| Model id | seedance-2.5-flash (seedance-2-5-flash and doubao-seedance-2-5-flash-261115 also work) |
| Clip length | 4 to 30 whole seconds, or duration: -1 to let the model choose (30 s reserved, the delivered length charged) |
| Resolution | 480p or 720p; no 1080p |
| Aspect ratio | 1:1, 4:3, 3:4, 16:9, 9:16, 21:9 or adaptive; adaptive only when there is a first frame |
| Inputs | A prompt; a first frame with an optional last frame; or references. Frames and references never share a request |
| Reference images | Up to 10 |
| Reference videos | Up to 5, each 2–30 s, 30 s in total |
| Reference audio | Up to 5 |
| Audio | Generated with the picture by default; generate_audio: false for a silent clip |
| Output | mp4 or mov |
| Real people | Through the asset library (asset://); raw portrait URLs are rejected and not charged |
| Task types | Generation, video editing (task_type edit) and video extension (task_type extend) |
| Result files | Kept for 30 days |
| Failed tasks | Not charged |
Seedance 2.5 Flash pricing: what one clip costs
Per second of output, the same clips on Seedance 2.5 alongside, and a second rate once a reference video goes in.
$0.364
Smallest Seedance 2.5 Flash request
4 seconds at 480p, $0.091 per second, sound included. The longest clip, 30 seconds at 720p, is $5.85.
| Clip length (no reference video) | Seedance 2.5 Flash 480p | Seedance 2.5 Flash 720p | Seedance 2.5 480p | Seedance 2.5 720p |
|---|---|---|---|---|
| Per second | $0.091 | $0.195 | $0.120 | $0.270 |
| 4 s | $0.364 | $0.78 | $0.48 | $1.08 |
| 5 s | $0.455 | $0.975 | $0.60 | $1.35 |
| 10 s | $0.91 | $1.95 | $1.20 | $2.70 |
| 15 s | $1.365 | $2.925 | $1.80 | $4.05 |
| 30 s | $2.73 | $5.85 | $3.60 | $8.10 |
| Per billable second with a reference video | $0.056 | $0.119 | $0.072 | $0.162 |
10 s at 720p with a 5 s reference video
$1.785
(5 + 10) = 15 billable s × $0.119; on Seedance 2.5, 15 × $0.162 = $2.43
6 s at 480p with clips of 3.2 s and 4.6 s
$0.84
Each clip rounds up: (4 + 5 + 6) = 15 billable s × $0.056
duration -1 at 720p, delivered at 12 s
$2.34
$5.85 (30 s) reserved at creation; 12 × $0.195 charged on delivery
Editing an 8 s clip at 720p
$1.904
Output keeps the source length: (8 + 8) = 16 billable s × $0.119
Extending an 8 s clip by 6 s at 480p
$0.784
(8 + 6) = 14 billable s × $0.056
Seedance 2.5 Flash runs at a lower rate per second than Seedance 2.5: about 24% less at 480p and 28% less at 720p. Reference video length is measured when the task is created, each clip rounded up to a whole second; edits and extensions use the same reference-video rate. With duration -1 the task reserves 30 seconds and is charged for the length actually delivered. Reference images and reference audio add nothing, turning audio off does not change the rate, and a Seedance 2.5 Flash task that fails is not charged.
What Seedance 2.5 Flash is used for
Seven common jobs, each with what you send, what comes back and what it is not for.
Batches of 9:16 social clips
Send one prompt per clip with aspect_ratio 9:16 and get back a clip with sound for each request. A 10-second 480p clip is $0.91, so 50 of them are $45.50.
Not for: deliverables that must be 1080p.
promptaspect_ratioDrafting a shot before the final render
Run each planned shot at 480p to choose the take, about $0.728 for an 8-second try, then render the chosen shot on seedance-2.5; both use the same field names.
Not for: shots that need more than 10 reference images, 5 videos or 5 audio clips; Seedance 2.5 takes 30, 10 and 10.
resolutionseedance-2.5Product animation from a first frame
Put a product photo in first_frame_url, optionally a closing shot in last_frame_url, and describe the motion. The clip opens on that exact image; 10 seconds at 720p is $1.95.
Not for: a product shot that also needs reference images in the same request, because frames and references do not mix.
first_frame_urllast_frame_urlQuick edits and extensions of a clip
Put the clip in reference_video_urls with task_type edit to change it as the prompt says, or extend to continue it. Editing an 8-second clip at 720p is (8 + 8) × $0.119 = $1.904.
Not for: source clips longer than 30 seconds, or 1080p results.
task_typereference_video_urlsA consenting real person via asset://
Register the portrait once in the asset library and pass its asset:// reference in first_frame_url or reference_image_urls. In a single test on 2026-11-17, a 480p 4-second clip from one portrait took about 61 seconds.
Not for: a raw portrait URL, or anyone who has not agreed to appear.
asset://Camera moves from a reference video
Add a reference clip in reference_video_urls and ask for the same camera movement on new content. A 10-second 720p clip with a 5-second reference is 15 billable seconds, $1.785.
Not for: reference footage longer than 30 seconds in total.
reference_video_urlsA music-led clip from reference audio
Add up to 5 tracks in reference_audio_urls and describe the picture; reference audio adds nothing to the rate, so 15 seconds at 480p is $1.365.
Not for: an audio-only request with no prompt; Flash needs a prompt unless there is a first frame.
reference_audio_urlsReal people in Seedance 2.5 Flash: the asset library route
Raw real-face uploads are refused for the whole 2.5 series; a registered asset:// portrait is the way in.
ByteDance's Seedance 2.5 documentation says the 2.5 series does not support directly uploading reference images or videos that contain real human faces, and Seedance 2.5 Flash is part of that series.
On APIMODELS there is a route for people who have agreed to appear: register the portrait once with POST /v1/assets (charged $0.01 only when registration succeeds) and you get an asset:// reference. Put it in first_frame_url to animate that person from the opening frame, or in reference_image_urls to borrow their face and outfit in a new scene. Seedance 2.5 uses the same asset library.
A raw portrait URL sent straight to seedance-2.5-flash is rejected when the task is created, and nothing is charged. Frames and references still cannot be mixed, so a portrait used as first_frame_url cannot sit next to reference images in the same request.
In a single test on 2026-11-17, a 480p 4-second Seedance 2.5 Flash clip from one asset-library portrait came back in about 61 seconds. That is one run, not a promise, and the portrait and the output still go through content review.
# 1) Register the portrait (once). Returns asset_url like asset://asset-2026...-xxxx
curl -X POST https://api.apimodels.app/v1/assets \
-H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" \
-d '{"url": "https://example.com/portrait.jpg", "asset_type": "Image", "group_id": 1}'
# 2) Use it as the first frame of a Seedance 2.5 Flash clip
curl -X POST https://api.apimodels.app/v1/video/generations \
-H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" \
-d '{
"model": "seedance-2.5-flash",
"prompt": "She turns to the camera, smiles and waves, soft morning light",
"first_frame_url": "asset://asset-2026...-xxxx",
"duration": 4,
"resolution": "480p"
}'Use real-person material only with people who have agreed to it. A group for the portraits is created once with POST /v1/assets/groups; the Seedance 2.5 docs walk through every field.
Seedance 2.5 Flash or Seedance 2.5?
One model generation and the same field names; the table shows what changes between our two Seedance 2.5 models.
| Seedance 2.5 Flash | Seedance 2.5 | |
|---|---|---|
| Model id | seedance-2.5-flash | seedance-2.5 |
| Per second, 480p / 720p | $0.091 / $0.195 | $0.120 / $0.270 |
| Per billable second with a reference video | $0.056 / $0.119 | $0.072 / $0.162 |
| 10 s at 720p | $1.95 | $2.70 |
| Reference images / videos / audio | 10 / 5 / 5 | 30 / 10 / 10 |
| Clip length | 4–30 s | 4–30 s |
| Automatic length (duration: -1) | Yes; 30 s reserved, the delivered length charged | Yes |
| Video editing and extension | Yes; source + output seconds at $0.056 / $0.119 | Yes; source + output seconds at $0.072 / $0.162 |
| Resolution | 480p / 720p | 480p / 720p |
| Real people | Asset library (asset://) | Plain URL or asset library (asset://) |
| ByteDance's positioning | About 2x faster; closest to Seedance 2.5 in quality; "not a complete replacement" | The full Seedance 2.5 model |
Pick Seedance 2.5 when a job needs more than 10 images, 5 videos or 5 audio clips, when you want the full ByteDance model rather than its faster tier for a final render, or when a real-person photo has to go in as a plain URL. For everything else, generation, edits, extensions and automatic length included, Seedance 2.5 Flash does the same job at a lower rate per second. Neither model outputs 1080p here; that is Seedance 2.0.
Seedance 2.5 Flash vs Gemini Omni 1.2 Flash vs Kling 4.0 Flash
Three fast video models on APIMODELS, compared on what each one can take and what it costs here.
| Seedance 2.5 Flash | Gemini Omni 1.2 Flash | Kling 4.0 Flash | |
|---|---|---|---|
| Model id | seedance-2.5-flash | gemini-omni-1.2-flash | kling-4.0-flash |
| Clip length | 4–30 s | 3–15 s (3–10 s with first and last frame) | 3–20 s |
| Resolution | 480p, 720p | 720p, 1080p | 720p |
| Aspect ratios | 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, adaptive | 16:9, 9:16 | auto, 16:9, 9:16, 1:1, 21:9 |
| References | 10 images, 5 videos, 5 audio clips | Up to 3 images | 10 images, 5 videos, 7 named subjects (15 in total) |
| Frames | First + optional last | First + optional last | First frame only |
| Audio | On by default, can be turned off | Synced audio on every clip | Sound on every clip, no silent mode |
| Real-person route | Asset library (asset://) | Not documented on its model page | Not documented on its model page |
| Per second at 720p | $0.195 | $0.045 | $0.094 |
| Per second at 480p | $0.091 | No 480p | No 480p |
Seedance 2.5 Flash is not the lowest rate per second in this group. Pick Gemini Omni 1.2 Flash when you need 1080p or the lowest 720p rate and 15 seconds with up to 3 reference images is enough. Pick Kling 4.0 Flash when you want named subjects called by @mention and up to 15 materials in one 720p clip of up to 20 seconds.
Pick Seedance 2.5 Flash when a take runs past 20 seconds, when reference audio should drive the sound, when you want a 480p tier or a silent clip, or when a consenting real person has to appear through the asset library.
How to call the Seedance 2.5 Flash API
From an API key to a finished clip in three steps.
Get an API key
Register on apimodels.app and generate a key in the console; the same key and USD balance work for Seedance 2.5 Flash and the rest of the catalog.
Create the task
POST /v1/video/generations with model seedance-2.5-flash, a prompt, duration, resolution and aspect_ratio, plus frames or references if you have them. A taskId comes back straight away.
Poll or wait for the callback
Poll GET /v1/video/generations?task_id= every 10 to 15 seconds until the state is completed, or pass callback_url and we POST the result to you. resultUrls holds the clip; files are kept for 30 days.
curl -X POST https://api.apimodels.app/v1/video/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "seedance-2.5-flash",
"prompt": "A street food vendor tosses noodles in a flaming wok at a night market, the crowd cheers, sizzling oil and chatter",
"duration": 8,
"resolution": "480p",
"aspect_ratio": "9:16",
"generate_audio": true
}'
# { "code": 200, "msg": "success", "data": { "taskId": "clxxx", "state": "pending" } }
curl "https://api.apimodels.app/v1/video/generations?task_id=TASK_ID" -H "Authorization: Bearer YOUR_API_KEY"
# { "code": 200, "data": { "state": "completed", "resultUrls": ["https://r2.apimodels.app/videos/clxxx.mp4"] } }Seedance 2.5 Flash API in Python
The same request with the requests library, polling until the Seedance 2.5 Flash task completes or fails.
import time, requests
API = "https://api.apimodels.app/v1/video/generations"
H = {"Authorization": "Bearer YOUR_API_KEY"}
task = requests.post(API, headers=H, json={
"model": "seedance-2.5-flash",
"prompt": "A paper boat drifts down a rain-soaked street at dusk, slow dolly follow",
"duration": 5,
"resolution": "720p", # 5 × $0.195 = $0.975
"aspect_ratio": "16:9",
}).json()["data"]["taskId"]
while True:
data = requests.get(API, headers=H, params={"task_id": task}).json()["data"]
if data["state"] in ("completed", "failed"):
print(data) # failed tasks are not charged
break
time.sleep(10)Seedance 2.5 Flash limits and what can fail
What the model does not do here, and why a request can still fail.
Not available on Flash
No 1080p, no web search tool and no seed parameter. Each of these returns a 400 before a task is created, so nothing is charged. Edit, extend and duration -1 requests may also occasionally return a 400 that is not charged; seedance-2.5 offers the same task types.
Content review
Prompts, reference media and outputs go through content review. A rejection fails the task, and a raw real-face upload is refused outright; use the asset library for consenting people.
Queue timeouts
A task can also time out while it waits in the upstream queue. Our two single runs on 2026-11-17 took about 70 and 61 seconds, but that is not a guaranteed turnaround.
Failed tasks are not charged
Success is not guaranteed, but you pay only for delivered clips: a failed Seedance 2.5 Flash task costs nothing, so adjust the input and try again.
Sources
Checked on 2026-11-17. Claims attributed to ByteDance on this page come from these pages; prices and measured runs are our own.
About Seedance 2.5 Flash
Seedance 2.5 Flash is a Video Generation API provided by ByteDance. Seedance 2.5 Flash is the faster tier of ByteDance's Seedance 2.5 video model: it turns a prompt, a first frame or up to 20 reference files into a 4–30 second clip at 480p or 720p with sound generated alongside the picture. ByteDance says Flash generates about twice as fast as Seedance 2.5 and describes it as the model whose results come closest to Seedance 2.5. On APIMODELS it is the model id seedance-2.5-flash on POST /v1/video/generations, from $0.091 per second at 480p, and failed tasks are not charged. Each clip is any whole length from 4 to 30 seconds, in 1:1, 4:3, 3:4, 16:9, 9:16, 21:9 or adaptive, delivered as mp4 or mov; sound is on by default and generate_audio: false turns it off. Text-to-video needs only a prompt. Image-to-video starts from first_frame_url, with last_frame_url as an optional end frame. Reference-to-video takes up to 10 images in reference_image_urls, up to 5 videos in reference_video_urls (each 2 to 30 seconds, 30 seconds in total) and up to 5 audio clips in reference_audio_urls. Frames and references cannot be combined in one request. ByteDance's documentation says the Seedance 2.5 series does not accept reference images or videos showing real human faces as direct uploads. On APIMODELS, real-person video goes through the same asset library as Seedance 2.5: register the portrait once, then put its asset:// reference in first_frame_url to animate that person, or in reference_image_urls to borrow the look; a raw portrait URL is rejected when the task is created and is not charged. Billing is per second of output: $0.091 at 480p and $0.195 at 720p, so a 5-second clip costs $0.455 at 480p and $0.975 at 720p, and a 10-second clip $0.91 and $1.95. When a request includes reference video the rate is $0.056 at 480p and $0.119 at 720p per billable second, where billable seconds are the total length of the uploaded reference videos, each clip rounded up to a whole second, plus the generated length: a 10-second 720p clip with a 5-second reference video is 15 billable seconds, $1.785. Reference images and audio add nothing. For comparison, Seedance 2.5 on APIMODELS is $0.120 per second at 480p and $0.270 at 720p ($0.072 and $0.162 with a reference video), so Flash runs at a lower rate per second, about 24% less at 480p and 28% less at 720p; a 10-second 720p clip is $1.95 on Flash and $2.70 on Seedance 2.5. Seedance 2.5 takes up to 30 images, 10 videos and 10 audio clips per job, against 10, 5 and 5 on Flash. Flash also takes automatic duration and the edit and extend task types. With duration: -1 the model picks a length from 4 to 30 seconds; the request reserves 30 seconds at creation ($5.85 at 720p, $2.73 at 480p) and is charged only for the delivered length, so a 720p clip that comes back 12 seconds long costs 12 × $0.195 = $2.34. task_type edit takes the source clip in reference_video_urls, keeps its length (duration -1 or omitted) and is billed at the reference-video rate on source plus output seconds: editing an 8-second clip at 720p is (8 + 8) × $0.119 = $1.904. task_type extend works the same way with a duration of 4 to 30 or -1: extending an 8-second clip by 6 seconds at 480p is (8 + 6) × $0.056 = $0.784. Edits and extensions need aspect_ratio adaptive or left out. These three may occasionally return a 400 that is not charged; seedance-2.5 offers them too. There is no 1080p, no seed parameter and no web search tool. Typical uses: short social clips in batches, drafts of a shot before the final render, quick edits and extensions of existing clips, animating a product or a character from a first frame, a consenting real person from the asset library, and new clips that follow the camera movement of a reference video or the music of a reference audio clip. Not every request succeeds: content review or upstream queuing can fail a task, and failed tasks are not charged. In single tests on 2026-11-17, a 480p 4-second text-to-video clip finished in about 70 seconds, and a 480p 4-second clip from one reference portrait in the asset library in about 61 seconds; those are single tests, not a guarantee. The API is asynchronous: the create call returns a task id, and you poll GET /v1/video/generations?task_id= until the state is completed, or pass callback_url. Through APIMODELS platform, you can access this model via a unified API with transparent pay-as-you-go pricing. Current pricing: 480p · per second: $0.091, 480p · 5s: $0.455, 480p · 10s: $0.91, 720p · per second: $0.195, 720p · 5s: $0.975, 720p · 10s: $1.95, Reference video · 480p · per billable second: $0.056, Reference video · 720p · per billable second: $0.119.
Related model: Seedance 2.5 — when one job needs up to 30 images, 10 videos and 10 audio clips, or a real-person photo passed as a plain URL — from $0.120/s
Related model: Seedance 2.0 — when you need 1080p — 480p / 720p / 1080p, 4-15 seconds, from $0.092/s
Related model: Seedance 2.5 Lite — for a fixed 15- or 30-second 720p clip from text or up to 9 reference images at a flat $2.50 per clip
Seedance 2.5 Flash API access compared: apimodels.app vs BytePlus vs OpenRouter
On apimodels.app, Seedance 2.5 Flash is $0.091 per second at 480p and $0.195 at 720p, and failed tasks are not charged; automatic duration, video editing and video extension are available. The table compares what else differs between the places you can call it. To be plain about the gaps: OpenRouter exposes an optional seed parameter and lets each request set provider routing preferences, while APIMODELS offers no seed on Flash; and Flash itself has no 1080p, since ByteDance lists it at 480p and 720p.
| Provider | Billing | Before the first call | Real-person photos | API | Payment |
|---|---|---|---|---|---|
| apimodels.app | Per output second; with a reference video, reference seconds + output seconds; failed tasks not charged; 4–30 s | Email or Google sign-up; no ID check, no prepaid minimum | Through the asset library (asset://); a raw portrait URL is rejected at creation and not charged | POST /v1/video/generations; async task, poll or callback_url webhook | Card, PayPal, Alipay, USDT |
| BytePlus (official, international) | Not stated | Not stated | Not accepted as a direct upload (rule for the whole 2.5 series) | Async task (POST /contents/generations/tasks), poll or callback_url | Not stated |
| OpenRouter | Failed generations not billed; other billing terms not on the model listing | Not stated | Not stated | Async job (POST /api/v1/videos), poll GET /api/v1/videos/{id} or callback_url | Not stated |
Checked on 17 November 2026 against the BytePlus Seedance 2.5 series tutorial and the OpenRouter Seedance 2.5 Flash listing. “Not stated” means those pages do not say.
Key Features
- 4-30 Seconds in One Pass -- Any whole length from 4 to 30 seconds from a single request, at 480p or 720p, in 1:1, 4:3, 3:4, 16:9, 9:16, 21:9 or adaptive, as mp4 or mov
- Native Synced Audio -- Sound is generated together with the picture by default. Pass generate_audio: false when you want a silent clip
- Text, Frames or up to 20 References -- A prompt alone, a first frame with an optional last frame, or up to 10 reference images, 5 videos (2-30 seconds each, 30 seconds in total) and 5 audio clips. Frames and references cannot be combined in one request
- Real Person via the Asset Library -- ByteDance does not take raw real-face uploads for the 2.5 series. Register the portrait once and pass its asset:// reference, in first_frame_url to animate that person or in reference_image_urls to borrow the look. Consented subjects only
- From $0.091 per Second -- 480p $0.091/s and 720p $0.195/s, against $0.120/s and $0.270/s for Seedance 2.5 on this site. Reference images and audio add nothing
- Reference Video Billing -- $0.056 (480p) or $0.119 (720p) per billable second: the total length of your reference videos, each clip rounded up to a whole second, plus the generated length
- Edit, Extend and Automatic Length -- task_type edit or extend on a source clip, billed on source + output seconds at the reference-video rate; duration -1 reserves 30 s and charges the delivered length
- Limits, Stated Plainly -- No 1080p, no seed and no web search. Content review or upstream queuing can fail a request, and edit, extend or -1 requests may occasionally return a 400; failed tasks are not charged
Use Cases
Marketing Videos
Quickly generate brand promotion videos for ad campaigns and social media marketing.
Social Media Content
Create compelling short-form video content for platforms like TikTok, Instagram, and YouTube.
Product Demos
Generate product feature demonstrations and tutorials to improve user conversion.
Educational Content
Produce course explanations, knowledge explainers, and training videos at an affordable price.
Why APIMODELS
- Unified API -- One API key to access all models, no need to register on multiple platforms
- Transparent Pricing -- Pay-as-you-go with per-call pricing published on every model page, ideal for indie developers and startups
- Instant Access -- Start right after signup; pay by card, PayPal, Alipay or USDT
- Full Documentation -- Detailed API docs with code examples in cURL, Python, and Node.js
Frequently Asked Questions
How to use Seedance 2.5 Flash API?
Sign up at APIMODELS, get your API key, and call our unified API endpoint. We provide detailed API documentation with code examples in cURL, Python, and Node.js.
What is the difference between APIMODELS and the official ByteDance API?
APIMODELS offers the same Seedance 2.5 Flash model through our aggregation platform. We provide a unified API interface so you do not need separate accounts for each provider - one API key to access all models.
What is Seedance 2.5 Flash?
Seedance 2.5 Flash is the faster tier of ByteDance's Seedance 2.5 video model: it turns a prompt, a first frame or up to 20 reference files into a 4–30 second clip at 480p or 720p with sound generated alongside the picture. ByteDance says it generates about twice as fast as Seedance 2.5 and comes closest to Seedance 2.5 in quality. On apimodels.app it is the model id seedance-2.5-flash, from $0.091 per second at 480p, and failed tasks are not charged. Typical uses are batches of short social clips, drafts of a shot before the final render, quick edits and extensions of existing clips, product animation from a first frame and a consenting real person through the asset library.
Is there a Seedance 2.5 Flash API?
Yes. On apimodels.app the Seedance 2.5 Flash API is the model id seedance-2.5-flash on POST https://api.apimodels.app/v1/video/generations, called with one API key that also covers every other model. The call is asynchronous: it returns a taskId, and you poll GET /v1/video/generations?task_id= or pass callback_url. For example, an 8-second 480p clip costs $0.728. A task can still fail on content review or an upstream queue timeout, and failed tasks are not charged.
When was Seedance 2.5 Flash released?
ByteDance announced Seedance 2.5 Flash on 12 November 2026 and said its API would open on 15 November 2026. It came as a second tier next to Seedance 2.5, which ByteDance had released earlier. On apimodels.app the model was added on 17 November 2026 as seedance-2.5-flash. The launch date and the API date are ByteDance's own, reported by IT之家 on the day of the announcement.
How much does Seedance 2.5 Flash cost per second?
$0.091 per second at 480p and $0.195 per second at 720p when the request has no reference video. That rate covers text-to-video, first and last frames, reference images and reference audio: a 5-second clip is $0.455 at 480p or $0.975 at 720p, 10 seconds $0.91 or $1.95, and the 30-second maximum $2.73 or $5.85. With a reference video the rate is $0.056 (480p) or $0.119 (720p) per billable second, where billable seconds are the reference video length, each clip rounded up, plus the generated length: 10 seconds at 720p with a 5-second reference is 15 billable seconds, $1.785. Failed tasks are not charged.
What is the difference between Seedance 2.5 Flash and Seedance 2.5?
Seedance 2.5 Flash has a lower rate per second and takes fewer references per job; both do text, frames, references, editing, extension and automatic length. On this site Seedance 2.5 is $0.120 per second at 480p and $0.270 at 720p, so Flash at $0.091 and $0.195 costs about 24% and 28% less; with a reference video it is $0.056 / $0.119 against $0.072 / $0.162. A 10-second 720p clip is $1.95 on Flash and $2.70 on Seedance 2.5. Flash takes up to 10 images, 5 videos and 5 audio clips, Seedance 2.5 up to 30, 10 and 10. ByteDance says Flash is about twice as fast and the closest to Seedance 2.5 in quality, while its Dreamina page says Flash is not a complete replacement for Seedance 2.5.
Does Seedance 2.5 Flash support 1080p, and how long can a clip be?
No, Seedance 2.5 Flash outputs 480p or 720p only, and a clip is any whole number of seconds from 4 to 30. For 1080p on this site, use Seedance 2.0. You can also pass duration: -1 and let the model pick a length from 4 to 30 seconds: 30 seconds are reserved when the task is created ($5.85 at 720p) and only the delivered length is charged, so a 12-second 720p result costs $2.34. aspect_ratio is 1:1, 4:3, 3:4, 16:9, 9:16, 21:9 or adaptive, output_format is mp4 or mov, and sound is generated by default; generate_audio: false gives a silent clip at the same price. The shortest clip, 4 seconds at 480p, is $0.364 and the longest, 30 seconds at 720p, is $5.85.
Does Seedance 2.5 Flash support video editing and extension?
Yes. Send task_type "edit" or "extend" with the source clip in reference_video_urls and a prompt that says what to change or how to continue. An edit keeps the source length (leave duration out or pass -1), an extension takes a duration of 4 to 30 seconds or -1, and both need aspect_ratio adaptive or left out. Both are billed at the reference-video rate on source seconds plus output seconds: editing an 8-second clip at 720p is (8 + 8) × $0.119 = $1.904, and extending an 8-second clip by 6 seconds at 480p is (8 + 6) × $0.056 = $0.784. Without task_type the model may still treat a request with a reference video as an edit or extension, and the bill follows the delivered clip the same way. These requests may occasionally return a 400 that is not charged; seedance-2.5 offers the same task types.
What inputs does one Seedance 2.5 Flash request take?
A prompt, a first frame with an optional last frame, or references: up to 10 images, 5 videos and 5 audio clips. Frames and references cannot share a request, so a first_frame_url and a reference_image_urls list in the same body return a 400. Each reference video must be 2 to 30 seconds and all of them together at most 30 seconds; a last frame needs a first frame; with a first frame the prompt is optional and aspect_ratio must be adaptive or left out. In the prompt, say what each reference is for, for example the face from one image and the camera movement from one video; add task_type edit or extend when the reference video is the clip to change.
Can I use photos of real people with Seedance 2.5 Flash?
Not as a raw photo: ByteDance rejects real faces in uploads for the whole 2.5 series. On APIMODELS you can, through the asset library: register the portrait once with POST /v1/assets, then put the asset:// reference in first_frame_url to animate that person, or in reference_image_urls to borrow their look and outfit. A raw portrait URL is rejected when the task is created and is not charged. In a single test on 2026-11-17, a 480p 4-second clip from one asset-library portrait came back in about 61 seconds. Use real-person material only with people who have agreed to it.
Is Seedance 2.5 Flash the same as Seedance 2.0 Fast?
No. Seedance 2.5 Flash is the fast tier of the newer Seedance 2.5 generation, while Seedance 2.0 Fast is the fast tier of Seedance 2.0. On APIMODELS, seedance-2.0-fast runs 4 to 15 seconds at 480p or 720p from $0.071 per second, with up to 9 images, 3 videos and 3 audio clips; seedance-2.5-flash runs 4 to 30 seconds from $0.091 per second, with up to 10 images, 5 videos and 5 audio clips. ByteDance says Seedance 2.5 Flash is ahead of Seedance 2.0 overall. Pick 2.0 Fast for clips of up to 15 seconds from $0.071 per second, and 2.5 Flash for longer takes or more references.
Seedance 2.5 Flash or Gemini Omni Flash: which should I use?
Use Gemini Omni 1.2 Flash for 1080p or the lowest 720p rate, and Seedance 2.5 Flash for takes longer than 15 seconds, more references, editing or extending an existing clip, or a real person through the asset library. On APIMODELS, Gemini Omni 1.2 Flash is $0.045 per second at 720p and $0.052 at 1080p, runs 3 to 15 seconds and takes up to 3 reference images; Seedance 2.5 Flash is $0.195 at 720p (or $0.091 at 480p), runs 4 to 30 seconds and takes 10 images, 5 videos and 5 audio clips. Gemini Omni 1.1 Flash adds 4K and video input, at 4 to 10 seconds. Neither Gemini page documents a real-person route.
How long does Seedance 2.5 Flash take, and can a request fail?
In our single tests on 2026-11-17, a 480p 4-second text-to-video clip finished in about 70 seconds and a 480p 4-second clip from one asset-library portrait in about 61 seconds. Those are single runs, not a guarantee. Requests can fail: content review can reject a prompt or a piece of media, and a task can time out in the upstream queue. Failed tasks are not charged, so you can adjust the input and try again; poll every 10 to 15 seconds or pass callback_url.
How do I get an API key and call the Seedance 2.5 Flash API?
Create an account at apimodels.app, generate an API key in the console, and POST to https://api.apimodels.app/v1/video/generations with the key as a Bearer token: curl -X POST https://api.apimodels.app/v1/video/generations -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"model": "seedance-2.5-flash", "prompt": "A street food vendor tosses noodles in a flaming wok while the crowd cheers", "duration": 8, "resolution": "480p", "aspect_ratio": "9:16"}'. The response holds a taskId: poll GET /v1/video/generations?task_id= until the state is completed, or pass callback_url, then download the file within 30 days. That 8-second 480p clip costs $0.728; the same request works from Python with requests.
Can I try Seedance 2.5 Flash for free?
Not entirely. Eligible new accounts get $0.10 of trial credit, but the smallest Seedance 2.5 Flash request, 4 seconds at 480p, costs $0.364, so a first test needs a top-up; top-ups start at $10. The playground on this page shows the price before you click Generate, including the billable seconds when you add a reference video, and failed tasks are not charged.
What can Seedance 2.5 Flash do?
Seedance 2.5 Flash supports: Text to Video, First + Last Frame, Reference Images, Videos and Audio, Video Edit + Extend, Real Person via asset://, 4-30s, 480p / 720p, Native Synced Audio, mp4 / mov, From $0.091/s. See the APIMODELS docs for full parameters and call examples.
Can I use the Seedance 2.5 Flash API from any country, including mainland China?
Yes. APIMODELS does not restrict access by country: the site and the API answer requests from anywhere, and apimodels.app and api.apimodels.app are reachable directly from mainland China without a proxy. You call Seedance 2.5 Flash with one APIMODELS key, so you do not need a separate account with the model's developer.
What payment methods are supported? Can I pay with PayPal or crypto?
Yes to both. You can top up with a card (Visa, Mastercard and others, through Stripe), PayPal, Alipay, or USDT on TRON (TRC20) or BNB Smart Chain (BEP20). Card, PayPal and Alipay top-ups are credited immediately; USDT is credited once the transfer confirms on-chain. Auto top-up is optional, and a $50 top-up adds a $1 bonus ($3 on a $100 top-up). The balance is in USD and billed per use, with no subscription.
Related models
Seedance 2.5 prompt examples
See all 95 →Prompts shared by their authors — copy and adapt them. Each one credits its author and links back to the original post.
Seven-shot bedroom candid — cat, phone call, doorbell
Montage, multi-shot candid observational footage. Do not use a single camera angle or continuous take. Handheld documentary style with the f…
by oggii_0
Handheld DV morning dog-walk vlog
CAMERA: Handheld DV 16mm daily vlog footage. The video MUST begin with her holding the camera at arm's length in selfie mode while stepping…
by maxxmalist
Thirty seconds of hand-to-hand combat in a rain-flooded metro maintenance hall
生成一段完整30秒、16:9横屏、24fps、写实电影级质感的现代近身格斗视频。使用我上传的两组人物参考图:第一名人物固定为"主角",第二名人物固定为"敌人"。严格继承参考图中两人的面部、年龄、发型、体型、身高比例、服装、鞋子、配饰和整体气质。全程不得交换身份,不得变脸、改变服装…
by lansenai
Need prompts?
We curate copy-ready prompt libraries — every entry shows its full text and a sample result, ready to adapt.
Related reading
How to get access, regional availability, and how this model compares with its alternatives.