{
"model": "seedance-2.5-flash",
"content": [
{
"type": "text",
"text": "Create a short video with seedance-2.5-flash: clear subject motion, realistic lighting, stable camera, and production-ready framing."
}
],
"resolution": "720p",
"ratio": "adaptive",
"duration": 5,
"generate_audio": true
}价格与计费
seedance-2.5-flash 价格与计费
下方价格来自当前模型的 Flatkey 定价数据,以及定价 API 返回的可见分组。
充值余额
图像、视频、音频和文本模型共用同一个 Flatkey 余额和 API Key。
- 模型类型
- 文生视频
- API 接入
- /v1/videos
- 计费依据
- 秒
- 时长
- 4–30s
- 画面比例
- adaptive · 16:9 · 4:3 · 1:1 · 3:4 · 9:16
Seedance 2.5 功能
支持文生视频、图生视频和参考素材输入,可按需生成音频,并调整时长、分辨率等输出设置。
文生视频与图生视频
可以从文字场景开始,也可以为已设计的主体、产品或分镜提供参考图。
参考素材与帧控制
请求可包含图片、视频和音频参考素材,并支持在参考驱动工作流中指定首帧和尾帧角色。
音视频生成
将音频生成作为明确的请求选项;不要假设所选路由未文档化的语言列表或音频行为。
时长与输出控制
提交任务前可选择 4–30 秒、480p 或 720p、自适应或支持的固定比例,以及音频设置。
对比
对比 Seedance 2.5 与 Seedance 2.0:文档化视频字段
此对比只记录已文档化的 Seedance 2.5 行为;Seedance 2.0 的未核实值不会被推测。
| 能力 | Seedance 2.0(未重新审计) | Seedance 2.5 |
|---|---|---|
| 片段时长 | 本次审计未核实 | 每次请求 4–30 秒 |
| 参考输入 | 本次审计未核实 | 总计最多 50 个:30 张图片、10 个视频和 10 个音频 |
| 帧控制 | 本次审计未核实 | 支持首帧和尾帧角色 |
| 音频 | 本次审计未核实 | 可在每次请求中启用音频生成 |
| 编辑工作流 | 本次审计未核实 | 参考驱动及首尾帧工作流 |
接入前先选对请求方式
核对支持的工作流、当前配置的实时费用,以及 Flatkey 实际使用的参数契约。
当目录提供固定单价时,根据当前 Playground 配置和实时目录价格计算。
- 当前配置
- 720p · 5s
- 实时单价
- from $0.09288 / 秒 · 4s min
- 本次预估
- $0.4644 $0.09288 × 5s
最终费用以实际提交的请求和账户分组为准。Token 或分层计费无法简化成一个固定预估值。
工作流与 Model ID
只使用标记为可用的工作流。Flatkey 对外保持一个模型 ID,由请求体选择输入方式。
文生视频
通过文本提示词和支持的输出参数生成视频。
- 模型 ID
seedance-2.5-flash- 接口
/v1/videos
图生视频
让一张或多张参考图动起来,并尽量保持主体与构图。
- 模型 ID
seedance-2.5-flash- 接口
/v1/videos
参考生视频
使用已文档化的图片、视频或音频参考素材指导结果。
- 模型 ID
seedance-2.5-flash- 接口
/v1/videos
视频编辑
在路由支持时编辑已有视频。
- 模型 ID
seedance-2.5-flash- 接口
/v1/videos
视频续写
在路由支持时续写已有视频。
- 模型 ID
seedance-2.5-flash- 接口
/v1/videos
参数兼容矩阵
以下取值与 Playground 和请求预览使用同一份模型配置。
resolution分辨率480p · 720p
默认值: 720pratio画面比例adaptive · 16:9 · 4:3 · 1:1 · 3:4 · 9:16
默认值: adaptiveduration时长4–30
默认值: 5generate_audio生成音频开 · 关
默认值: 开image参考素材上限30 个文件
接口: /v1/videosvideo参考素材上限10 个文件
接口: /v1/videosaudio参考素材上限10 个文件
接口: /v1/videos从首次测试到生产接入
Playground 与 API 使用同一套请求契约,并明确处理参数校验、余额、任务状态和产物获取。
请求生命周期
- 1
校验模型、提示词、参考素材和输出设置。
- 2
提交 POST 请求并保存返回的任务 ID。
- 3
轮询任务状态或处理已配置的回调。
- 4
从产物接口获取文件,并在失效前保存到自己的存储。
生产检查项
提交前拦截不支持的尺寸、时长、比例、格式或参考素材数量。
分别处理无效密钥、余额不足、限流和账户分组价格。
明确指出被拒绝的提示词或素材,让用户只替换失败的输入。
保留任务 ID,限制重试次数,且不能把超时当成生成成功。
提示词库
Seedance 2.5 提示词指南:六种工作流
这些按工作流编写的提示词可作为起点;输出取决于参考素材和请求设置。
Create a 9:16 micro-drama storyboard from a three-beat comic script: a courier notices a torn envelope, follows a red umbrella through a crowded station, and stops at a silent platform. Keep the same two characters, wardrobe, props, and rainy evening light across the beats; use clear eyelines and no readable text.
收起提示词
Create a 9:16 micro-drama storyboard from a three-beat comic script: a courier notices a torn envelope, follows a red umbrella through a crowded station, and stops at a silent platform. Keep the same two characters, wardrobe, props, and rainy evening light across the beats; use clear eyelines and no readable text.
Create a 9:16 e-commerce UGC video for a reusable travel bottle. Show an adult creator opening the package, filling the bottle, and placing it in a tote bag. Keep the bottle shape and lid color consistent, use a small apartment kitchen, leave space for Portuguese captions added in post, and do not invent logos or readable label text.
收起提示词
Create a 9:16 e-commerce UGC video for a reusable travel bottle. Show an adult creator opening the package, filling the bottle, and placing it in a tote bag. Keep the bottle shape and lid color consistent, use a small apartment kitchen, leave space for Portuguese captions added in post, and do not invent logos or readable label text.
Create a 12-second film previsualization of a detective entering an empty observatory at dawn. Start with a wide establishing view, track behind the character, then pan to the telescope and hold on the doorway. Keep screen direction, blocking, and the warm-to-cool light transition consistent; no dialogue text or logos.
收起提示词
Create a 12-second film previsualization of a detective entering an empty observatory at dawn. Start with a wide establishing view, track behind the character, then pan to the telescope and hold on the doorway. Keep screen direction, blocking, and the warm-to-cool light transition consistent; no dialogue text or logos.

Animate a game-cinematic reference board into an 8-second shot: a masked pilot walks across a hangar while a grounded shuttle powers up behind them. Preserve the pilot silhouette, helmet markings, shuttle geometry, and camera-left-to-right movement; use practical hangar lights and restrained smoke, with no new text or logos.
收起提示词
Animate a game-cinematic reference board into an 8-second shot: a masked pilot walks across a hangar while a grounded shuttle powers up behind them. Preserve the pilot silhouette, helmet markings, shuttle geometry, and camera-left-to-right movement; use practical hangar lights and restrained smoke, with no new text or logos.

Create a 4:5 creator explainer about a compact microphone for an independent video maker. Begin with a talking-head medium shot, cut to a close-up of the cable connection, then return to the same framing for a practical tip. Keep the speaker, microphone, and room layout consistent; leave clean caption space and avoid readable brand text.
收起提示词
Create a 4:5 creator explainer about a compact microphone for an independent video maker. Begin with a talking-head medium shot, cut to a close-up of the cable connection, then return to the same framing for a practical tip. Keep the speaker, microphone, and room layout consistent; leave clean caption space and avoid readable brand text.
Create three short, clearly distinct creative variants for a market-research test of a sunscreen product: beach morning, city commute, and family picnic. Keep the same bottle proportions and cap, change only setting and opening action, use neutral props, and leave all claims and captions for post-production. Do not add medical promises or invented labels.
收起提示词
Create three short, clearly distinct creative variants for a market-research test of a sunscreen product: beach morning, city commute, and family picnic. Keep the same bottle proportions and cap, change only setting and opening action, use neutral props, and leave all claims and captions for post-production. Do not add medical promises or invented labels.
为什么选择 Flatkey
为什么通过 Flatkey 使用 Seedance 2.5?
一个 Key 覆盖模型目录,清楚查看请求契约,并让价格与所选设置保持一致。
一个 Key 覆盖模型目录
视频、图片、音频和文本任务共用同一个 Flatkey 账号与 API Key。
文档化视频契约
在集成中明确 Seedance 的 content[] 请求、/v1/videos endpoint 和异步任务流程。
价格随请求变化
分辨率、时长和视频参考输入会改变公式;目录值不是统一的每秒价格承诺。
仅在有数据时显示实时指标
只有积累足够真实 Flatkey 流量时才显示遥测卡片,否则保持未报告状态。
API 接入
seedance-2.5-flash API
四种接入方式共用同一个 Key 和模型目录,选择一种查看可运行示例。
curl https://router.flatkey.ai/v1/videos \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-***" \ -d '{"model":"seedance-2.5-flash","content":[{"type":"text","text":"Create a short video with seedance-2.5-flash: clear subject motion, realistic lighting, stable camera, and production-ready framing."}],"ratio":"adaptive","resolution":"720p","duration":5,"generate_audio":true}' # Poll for the result (replace TASK_ID with the returned task id) curl "https://router.flatkey.ai/v1/videos/TASK_ID" \ -H "Authorization: Bearer sk-***"
常见问题
Seedance 2.5
API、价格和发布日期常见问题
价格、兼容性、限制,以及提示词和生成文件的处理方式。