Usage docs: API integration

llmgate is compatible with OpenAI-style `/v1` calls. Your app only needs to replace the API Key and base_url while continuing to use the public platform model names.

API endpoint

https://api.starchasing.ai/v1

Text example model

glm-5

Image example model

doubao-seedream-5-0-lite

Integration

API integration flow

Create API Key Export MD

管理员创建账号并充值

After an administrator creates the account, recharge the wallet. All API Keys share the same user wallet.

Create API Key

Generate sk-lg-* in the dashboard. Your app uses the key for unified authentication.

Replace base_url

Use https://api.starchasing.ai/v1 as the API endpoint.

Send a test request

Send one minimal request using a public model name, then verify requestId, tokens, and charges in dashboard usage logs.

Shared

Common supported API endpoints

These endpoints are shared across model types for integration, model discovery, and usage review. Text, image, and video endpoints are placed under their model sections.

MethodPathAuthUse

GET/v1/modelsNo API Key requiredReturns model list, callable status, capabilities, product channels, and current prices for pre-integration checks.

GET/v1/usageBearer sk-lg-*Lists usage logs for the current account, filterable by time, request_id, model, status, and API Key.

GET/v1/usage/{request_id}Bearer sk-lg-*Gets tokens, charge, status, user-visible error info, and sync-video deferred timing for one request_id.

Text Models

Text model calls and pricing

Text models all use OpenAI-compatible Chat Completions. Each model card carries its own price, context, and tiering notes. Overseas prices use official USD sources when the vendor publishes them; models without a current official USD row are marked instead of being converted from CNY.

GLM

glm-5

POST /v1/chat/completions

GLM text, coding, and general QA model for chat, summarization, structured extraction, and intent understanding before tool calls.

Model price

per 1M tokens

input $1.00 / output $3.20 / cache read $0.20

Tier basis: input_context_tokens

Z.ai overseas official USD pricing.Official pricing

Model endpoint

POST/v1/chat/completions

For text models, only replace the `model` field; the same endpoint calls all text models listed here.

ParameterDescription

model

Required. Use the public model name on this card, for example glm-5. Do not send upstream-internal model IDs.

messages

Required OpenAI Chat Completions message array. For text models, use string content. Do not copy multimodal inputs into these text models unless /v1/models shows the capability.

stream

Optional boolean. true returns an SSE stream; non-streaming returns one JSON response. Both modes record tokens, charges, and requestId in usage logs.

max_tokens

Optional output-token limit for this request; it is not the context window. The context limit follows /v1/models and the current route.

temperature / top_p

Optional sampling parameters. Omit them unless you have a specific need. If a route cannot support a specific value, the platform handles it according to route capability or falls back.

tools / tool_choice

Optional function-calling parameters. Use them only when the model and current route are confirmed to support tool calls; omit them otherwise.

response_format

Optional JSON output control. Verify the model with a normal text request before enabling JSON mode; strict adherence depends on model and route capability.

The request model must be glm-5; do not use vendor aliases or upstream-internal IDs.

Responses and dashboard usage logs keep the platform requestId, input/output tokens, cache-read tokens, charged amount, and product-channel snapshot.

The overseas docs use the current public Z.ai USD pricing. Do not convert BigModel CNY pricing into an overseas official price.

Minimal call example

curl https://api.starchasing.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5",
    "messages": [
      {"role": "system", "content": "You are a concise API integration assistant."},
      {"role": "user", "content": "Summarize three risks to check before switching an app to a new API gateway."}
    ],
    "stream": false,
    "temperature": 0.7,
    "max_tokens": 800
  }'

GLM

glm-5.1

POST /v1/chat/completions

GLM 5.1 text model for heavier reasoning, coding, and longer-input tasks than glm-5.

Model price

per 1M tokens

input $1.40 / output $4.40 / cache read $0.26

Tier basis: input_context_tokens

Z.ai overseas official USD pricing.Official pricing

Model endpoint

POST/v1/chat/completions

For text models, only replace the `model` field; the same endpoint calls all text models listed here.

ParameterDescription

model

Required. Use the public model name on this card, for example glm-5. Do not send upstream-internal model IDs.

messages

Required OpenAI Chat Completions message array. For text models, use string content. Do not copy multimodal inputs into these text models unless /v1/models shows the capability.

stream

Optional boolean. true returns an SSE stream; non-streaming returns one JSON response. Both modes record tokens, charges, and requestId in usage logs.

max_tokens

Optional output-token limit for this request; it is not the context window. The context limit follows /v1/models and the current route.

temperature / top_p

Optional sampling parameters. Omit them unless you have a specific need. If a route cannot support a specific value, the platform handles it according to route capability or falls back.

tools / tool_choice

Optional function-calling parameters. Use them only when the model and current route are confirmed to support tool calls; omit them otherwise.

response_format

Optional JSON output control. Verify the model with a normal text request before enabling JSON mode; strict adherence depends on model and route capability.

The request model must be glm-5.1; do not use vendor aliases or upstream-internal IDs.

Responses and dashboard usage logs keep the platform requestId, input/output tokens, cache-read tokens, charged amount, and product-channel snapshot.

The overseas docs use the current public Z.ai USD pricing. Do not convert BigModel CNY pricing into an overseas official price.

Minimal call example

curl https://api.starchasing.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.1",
    "messages": [
      {"role": "system", "content": "You are a concise API integration assistant."},
      {"role": "user", "content": "Summarize three risks to check before switching an app to a new API gateway."}
    ],
    "stream": false,
    "temperature": 0.7,
    "max_tokens": 800
  }'

GLM

glm-5.2

POST /v1/chat/completions

GLM 5.2 text model exposed through the same Chat Completions endpoint on the overseas site; the context limit follows the model list and current route.

Model price

per 1M tokens

input $1.40 / output $4.40 / cache read $0.26

Z.ai overseas official USD pricing.Official pricing

Model endpoint

POST/v1/chat/completions

For text models, only replace the `model` field; the same endpoint calls all text models listed here.

ParameterDescription

model

Required. Use the public model name on this card, for example glm-5. Do not send upstream-internal model IDs.

messages

Required OpenAI Chat Completions message array. For text models, use string content. Do not copy multimodal inputs into these text models unless /v1/models shows the capability.

stream

Optional boolean. true returns an SSE stream; non-streaming returns one JSON response. Both modes record tokens, charges, and requestId in usage logs.

max_tokens

Optional output-token limit for this request; it is not the context window. The context limit follows /v1/models and the current route.

temperature / top_p

Optional sampling parameters. Omit them unless you have a specific need. If a route cannot support a specific value, the platform handles it according to route capability or falls back.

tools / tool_choice

Optional function-calling parameters. Use them only when the model and current route are confirmed to support tool calls; omit them otherwise.

response_format

Optional JSON output control. Verify the model with a normal text request before enabling JSON mode; strict adherence depends on model and route capability.

The request model must be glm-5.2; do not use vendor aliases or upstream-internal IDs.

Responses and dashboard usage logs keep the platform requestId, input/output tokens, cache-read tokens, charged amount, and product-channel snapshot.

The overseas docs use the current public Z.ai USD pricing. Do not convert BigModel CNY pricing into an overseas official price.

External GLM-5.2 docs may mention a higher context window. Do not hard-code the context limit in business code; read /v1/models and the dashboard model list before integration.

Minimal call example

curl https://api.starchasing.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "messages": [
      {"role": "system", "content": "You are a concise API integration assistant."},
      {"role": "user", "content": "Summarize three risks to check before switching an app to a new API gateway."}
    ],
    "stream": false,
    "temperature": 0.7,
    "max_tokens": 800
  }'

DeepSeek

deepseek-v3.2

POST /v1/chat/completions

General DeepSeek text model for chat, long-form summarization, coding assistance, and structured output.

Model price

per 1M tokens

The current DeepSeek pricing page does not publish a numeric USD row for deepseek-v3.2. Do not convert CNY for this model; billing follows the dashboard request snapshot.

DeepSeek V3.2 release and current DeepSeek pricing page.Release note Official pricing

Model endpoint

POST/v1/chat/completions

For text models, only replace the `model` field; the same endpoint calls all text models listed here.

ParameterDescription

model

Required. Use the public model name on this card, for example glm-5. Do not send upstream-internal model IDs.

messages

Required OpenAI Chat Completions message array. For text models, use string content. Do not copy multimodal inputs into these text models unless /v1/models shows the capability.

stream

Optional boolean. true returns an SSE stream; non-streaming returns one JSON response. Both modes record tokens, charges, and requestId in usage logs.

max_tokens

Optional output-token limit for this request; it is not the context window. The context limit follows /v1/models and the current route.

temperature / top_p

Optional sampling parameters. Omit them unless you have a specific need. If a route cannot support a specific value, the platform handles it according to route capability or falls back.

tools / tool_choice

Optional function-calling parameters. Use them only when the model and current route are confirmed to support tool calls; omit them otherwise.

response_format

Optional JSON output control. Verify the model with a normal text request before enabling JSON mode; strict adherence depends on model and route capability.

The request model must be deepseek-v3.2; do not use vendor aliases or upstream-internal IDs.

Responses and dashboard usage logs keep the platform requestId, input/output tokens, cache-read tokens, charged amount, and product-channel snapshot.

This model is currently shown with flat token pricing and no 32K tier. Any price-group override is determined by the request's price snapshot.

Minimal call example

curl https://api.starchasing.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v3.2",
    "messages": [
      {"role": "system", "content": "You are a concise API integration assistant."},
      {"role": "user", "content": "Summarize three risks to check before switching an app to a new API gateway."}
    ],
    "stream": false,
    "temperature": 0.7,
    "max_tokens": 800
  }'

DeepSeek

deepseek-v4-flash

POST /v1/chat/completions

DeepSeek Flash text model optimized for speed and cost, suitable for classification, extraction, lightweight QA, and batch jobs.

Model price

per 1M tokens

input $0.14 / output $0.28 / cache read $0.01

DeepSeek overseas official USD pricing.Official pricing

Model endpoint

POST/v1/chat/completions

For text models, only replace the `model` field; the same endpoint calls all text models listed here.

ParameterDescription

model

Required. Use the public model name on this card, for example glm-5. Do not send upstream-internal model IDs.

messages

Required OpenAI Chat Completions message array. For text models, use string content. Do not copy multimodal inputs into these text models unless /v1/models shows the capability.

stream

Optional boolean. true returns an SSE stream; non-streaming returns one JSON response. Both modes record tokens, charges, and requestId in usage logs.

max_tokens

Optional output-token limit for this request; it is not the context window. The context limit follows /v1/models and the current route.

temperature / top_p

Optional sampling parameters. Omit them unless you have a specific need. If a route cannot support a specific value, the platform handles it according to route capability or falls back.

tools / tool_choice

Optional function-calling parameters. Use them only when the model and current route are confirmed to support tool calls; omit them otherwise.

response_format

Optional JSON output control. Verify the model with a normal text request before enabling JSON mode; strict adherence depends on model and route capability.

The request model must be deepseek-v4-flash; do not use vendor aliases or upstream-internal IDs.

Responses and dashboard usage logs keep the platform requestId, input/output tokens, cache-read tokens, charged amount, and product-channel snapshot.

This model is currently shown with flat token pricing and no 32K tier. Any price-group override is determined by the request's price snapshot.

Minimal call example

curl https://api.starchasing.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [
      {"role": "system", "content": "You are a concise API integration assistant."},
      {"role": "user", "content": "Summarize three risks to check before switching an app to a new API gateway."}
    ],
    "stream": false,
    "temperature": 0.7,
    "max_tokens": 800
  }'

DeepSeek

deepseek-v4-pro

POST /v1/chat/completions

DeepSeek Pro text model for more complex reasoning, coding, and higher-quality generation tasks.

Model price

per 1M tokens

input $0.44 / output $0.87 / cache read $0.01

DeepSeek overseas official USD pricing.Official pricing

Model endpoint

POST/v1/chat/completions

For text models, only replace the `model` field; the same endpoint calls all text models listed here.

ParameterDescription

model

Required. Use the public model name on this card, for example glm-5. Do not send upstream-internal model IDs.

messages

Required OpenAI Chat Completions message array. For text models, use string content. Do not copy multimodal inputs into these text models unless /v1/models shows the capability.

stream

Optional boolean. true returns an SSE stream; non-streaming returns one JSON response. Both modes record tokens, charges, and requestId in usage logs.

max_tokens

Optional output-token limit for this request; it is not the context window. The context limit follows /v1/models and the current route.

temperature / top_p

Optional sampling parameters. Omit them unless you have a specific need. If a route cannot support a specific value, the platform handles it according to route capability or falls back.

tools / tool_choice

Optional function-calling parameters. Use them only when the model and current route are confirmed to support tool calls; omit them otherwise.

response_format

Optional JSON output control. Verify the model with a normal text request before enabling JSON mode; strict adherence depends on model and route capability.

The request model must be deepseek-v4-pro; do not use vendor aliases or upstream-internal IDs.

Responses and dashboard usage logs keep the platform requestId, input/output tokens, cache-read tokens, charged amount, and product-channel snapshot.

This model is currently shown with flat token pricing and no 32K tier. Any price-group override is determined by the request's price snapshot.

Minimal call example

curl https://api.starchasing.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "messages": [
      {"role": "system", "content": "You are a concise API integration assistant."},
      {"role": "user", "content": "Summarize three risks to check before switching an app to a new API gateway."}
    ],
    "stream": false,
    "temperature": 0.7,
    "max_tokens": 800
  }'

Image Generation

Image generation models

Image parameters are model-specific. Do not mix parameters between models. Sync and async image generation use the same JSON body.

Doubao Seedream 5.0 Lite

doubao-seedream-5-0-lite

/v1/images/generations/v1/images/generations/async

Public image model for text-to-image and reference-image generation. The current platform standard channel is billed per image.

Model price

per request

$0.04 / image

BytePlus ModelArk overseas official USD pricing.Official pricing

ParameterDescription

model

Required. Use the public model name shown in this section.

prompt

Required image generation prompt as a string.

size

Optional. Accepts a size tier, aspect ratio, or pixel dimensions; the platform normalizes it to the model's official size rules before upstream submission.

resolution

Optional platform compatibility field. When present it decides the size tier; otherwise the platform infers the tier from size.

image / reference_images / images

Optional reference images. Accepts URLs or data URLs; image is convenient for one reference image, and array fields are convenient for multiple images.

use_reference_image_size

Optional boolean. Seedream image generations only; an explicit non-empty size takes priority. When size is omitted and a reference image is present, the platform uses the first data URL or public URL reference image dimensions as the size source. Missing, unsafe, or unreadable reference dimensions return HTTP 400.

n

Optional number of images to generate; billing follows generated image count.

response_format

Optional. Common values are url and b64_json; the actual response format depends on the current route capability.

watermark

Optional boolean for AI-generated watermarking; set false to generate without the watermark.

size / resolution values

Supports 2K, 3K, and 4K, plus aspect ratios such as 1:1, 16:9, 9:16, or WIDTHxHEIGHT. 1K is normalized to a Seedream-compatible size by the platform.

When neither size nor resolution is provided, the platform defaults to 2K.

Custom dimensions must normalize into the model's accepted range. Unsupported size or resolution returns HTTP 400 and is not charged.

Official parameters not exposed or not verified on current routes

Official parameterCurrent route status

stream

Official docs include streaming capability. The current public Seedream routes return sync results or async tasks and do not guarantee image streaming.

sequential_image_generation

Official docs include sequential image generation. The current public Seedream image routes have not been verified for it, so it is not exposed as a user-facing parameter.

sequential_image_generation_options

Official docs include options for sequential image generation. They depend on sequential generation, which is not exposed on the current public routes.

tools

Official docs include tool capability such as web search. The current public Seedream image routes have not been verified for it and do not guarantee it works.

optimize_prompt_options

Official docs include prompt optimization options. The current public routes have not been verified for them and do not expose them as user-facing contract.

Seedream 5.0 Lite image generation

curl https://api.starchasing.ai/v1/images/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-5-0-lite",
    "prompt": "A clean product photo of a ceramic coffee cup",
    "size": "3K",
    "image": "https://example.com/reference.png",
    "n": 1,
    "response_format": "url"
  }'

Doubao Seedream 4.5

doubao-seedream-4-5

/v1/images/generations/v1/images/generations/async

Public image model for high-quality text-to-image and reference-image generation. The current platform standard channel is billed per image.

Model price

per request

$0.04 / image

BytePlus ModelArk overseas official USD pricing.Official pricing

ParameterDescription

model

Required. Use the public model name shown in this section.

prompt

Required image generation prompt as a string.

size

Optional. Accepts a size tier, aspect ratio, or pixel dimensions; the platform normalizes it to the model's official size rules before upstream submission.

resolution

Optional platform compatibility field. When present it decides the size tier; otherwise the platform infers the tier from size.

image / reference_images / images

Optional reference images. Accepts URLs or data URLs; image is convenient for one reference image, and array fields are convenient for multiple images.

use_reference_image_size

n

Optional number of images to generate; billing follows generated image count.

response_format

Optional. Common values are url and b64_json; the actual response format depends on the current route capability.

watermark

Optional boolean for AI-generated watermarking; set false to generate without the watermark.

size / resolution values

Supports 2K and 4K, plus aspect ratios such as 1:1, 16:9, 9:16, or WIDTHxHEIGHT. 3K is not a public supported tier for this model.

When neither size nor resolution is provided, the platform defaults to 2K.

doubao-seedream-4-5 does not document 3K, so clients do not copy another model's capability by mistake.

Official parameters not exposed or not verified on current routes

Official parameterCurrent route status

stream

Official docs include streaming capability. The current public Seedream routes return sync results or async tasks and do not guarantee image streaming.

sequential_image_generation

Official docs include sequential image generation. The current public Seedream image routes have not been verified for it, so it is not exposed as a user-facing parameter.

sequential_image_generation_options

Official docs include options for sequential image generation. They depend on sequential generation, which is not exposed on the current public routes.

tools

Official docs include tool capability such as web search. The current public Seedream image routes have not been verified for it and do not guarantee it works.

optimize_prompt_options

Official docs include prompt optimization options. The current public routes have not been verified for them and do not expose them as user-facing contract.

Seedream 4.5 image generation

curl https://api.starchasing.ai/v1/images/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-4-5",
    "prompt": "A clean product photo of a ceramic coffee cup",
    "size": "4K",
    "image": "https://example.com/reference.png",
    "n": 1,
    "response_format": "url"
  }'

Video Generation

Video generation models

Video parameters are also model-specific. Use the async endpoint for long-running jobs, then poll task status.

Doubao Seedance 2.0

doubao-seedance-2-0

/v1/videos/generations/v1/videos/generations/async

Public video model for text-to-video, image-to-video, and multimodal references. Use the async endpoint for long-running jobs.

Model price

per 1M output tokens

480p/720p: input without video $7.00 / input with video $4.30

1080p: input without video $7.70 / input with video $4.70

4K: input without video $4.00 / input with video $2.40

BytePlus ModelArk overseas official USD pricing.Official pricing

ParameterDescription

model

Required. Use the public model name shown in this section.

content

Required official content array. Include at least one text prompt; reference image, video, and audio use image_url, video_url, and audio_url blocks.

duration

Optional video duration. Accepted values depend on the model and current upstream capability.

ratio

Optional Seedance aspect ratio enum: 16:9, 4:3, 1:1, 3:4, 9:16, 21:9, or adaptive. For Seedance 2.0 / 2.0 fast, when omitted, the platform fills the official default adaptive before forwarding upstream, so the model chooses the aspect ratio from the reference media and input content. Use a fixed enum only when you need to force a specific aspect ratio.

resolution

Optional official Seedance resolution field. Use 480p, 720p, 1080p, or 4k; omitted requests are billed as 720p. The platform also accepts video_resolution, videoResolution, and size as resolution aliases, but new integrations should use resolution.

generate_audio

Optional boolean for whether to generate audio.

watermark

Optional boolean for watermarking.

seed

Optional integer random seed.

return_last_frame

Optional boolean for returning the last frame.

tools

Optional. Pass [{"type":"web_search"}] only when web search is needed.

Use role: "first_frame" / role: "last_frame" for frame references; use role: "reference_image" for normal image references. A last frame cannot be used alone and requires a first frame.

The sync endpoint may return media directly or an upstream task. For async video routes such as Seedance and HappyHorse, the platform normalizes provider-specific envelopes into top-level id, task_id, status, and video_urls. Whenever a pollable task id is present, the response is status=running and an internal settlement task is created; even if upstream returns usage at submission time, the user is not charged as success until deliverable media is available. During submission, an async video upstream HTTP 200 without both a task id and a video URL is recorded as UPSTREAM_VIDEO_RESULT_MISSING for the current route, then the request continues through fallback; only when every candidate is undeliverable is the request returned as failed. Known in-progress sync video tasks stay status=running on temporary upstream polling 5xx responses; after the task is stored locally, polling returns the platform-hosted video URL first. The platform async endpoint lets workers handle polling, storage, and settlement.

Official parameters not exposed or not verified on current routes

Official parameterCurrent route status

callback_url

Official docs include callback URLs. Public platform async tasks do not expose customer callbacks; platform workers poll, store media, settle billing, and clients query /tasks/{taskId}.

Seedance 2.0 video generation

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0",
    "content": [
      {"type": "text", "text": "A short cinematic shot of a coffee cup on a wooden desk"},
      {"type": "image_url", "image_url": {"url": "https://example.com/reference.png"}, "role": "reference_image"}
    ],
    "duration": 5,
    "ratio": "16:9",
    "resolution": "1080p",
    "generate_audio": true,
    "watermark": false
  }'

Seedance 2.0 first/last frame video generation

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0",
    "content": [
      {"type": "text", "text": "Make a smooth cinematic transition from the first image to the second image"},
      {"type": "image_url", "role": "first_frame", "image_url": {"url": "https://example.com/first-frame.png"}},
      {"type": "image_url", "role": "last_frame", "image_url": {"url": "https://example.com/last-frame.png"}}
    ],
    "duration": 5,
    "ratio": "16:9",
    "resolution": "1080p",
    "generate_audio": false,
    "watermark": false,
    "return_last_frame": true
  }'

Doubao Seedance 2.0 Fast

doubao-seedance-2-0-fast

/v1/videos/generations/v1/videos/generations/async

Public video model optimized for speed. It uses the same official content-array request shape as Seedance 2.0.

Model price

per 1M output tokens

input without video: $5.60 / input with video: $3.30

BytePlus ModelArk overseas official USD pricing.Official pricing

ParameterDescription

model

Required. Use the public model name shown in this section.

content

Required official content array. Include at least one text prompt; reference image, video, and audio use image_url, video_url, and audio_url blocks.

duration

Optional video duration. Accepted values depend on the model and current upstream capability.

ratio

resolution

generate_audio

Optional boolean for whether to generate audio.

watermark

Optional boolean for watermarking.

seed

Optional integer random seed.

return_last_frame

Optional boolean for returning the last frame.

Do not copy the image-model prompt + image shape for the Fast model; video requests use the content array. First/last frame requests use the same first_frame / last_frame roles.

The platform currently bills video generation by output tokens. Creation returns task state first; final settlement happens after completion.

Official parameters not exposed or not verified on current routes

Official parameterCurrent route status

callback_url

Official docs include callback URLs. Public platform async tasks do not expose customer callbacks; platform workers poll, store media, settle billing, and clients query /tasks/{taskId}.

tools

The same official family includes web-search tool syntax. The Fast route has not been verified for it, so it is not listed as public supported parameter. Use doubao-seedance-2-0 when web search is needed.

Seedance 2.0 Fast video generation

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0-fast",
    "content": [
      {"type": "text", "text": "A short cinematic shot of a coffee cup on a wooden desk"},
      {"type": "image_url", "image_url": {"url": "https://example.com/reference.png"}, "role": "reference_image"}
    ],
    "duration": 5,
    "ratio": "16:9",
    "resolution": "720p",
    "generate_audio": true,
    "watermark": false
  }'

Seedance 2.0 Fast first/last frame video generation

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0-fast",
    "content": [
      {"type": "text", "text": "Make a smooth cinematic transition from the first image to the second image"},
      {"type": "image_url", "role": "first_frame", "image_url": {"url": "https://example.com/first-frame.png"}},
      {"type": "image_url", "role": "last_frame", "image_url": {"url": "https://example.com/last-frame.png"}}
    ],
    "duration": 5,
    "ratio": "16:9",
    "resolution": "720p",
    "generate_audio": false,
    "watermark": false,
    "return_last_frame": true
  }'

HappyHorse

HappyHorse video generation manual

This is the StarChasing client-facing request shape, not the raw Alibaba Cloud request envelope. Clients call StarChasing `/v1/videos/generations` or the async task APIs; the platform handles Alibaba Cloud's required async header, task polling, and media storage.

Task Flow

Create a task, then poll the result

The sync endpoint POST /v1/videos/generations creates a HappyHorse task; when the response contains task_id, treat it as status=running first.

Poll sync calls with GET /v1/videos/generations/{task_id}; the same request's requestId is also accepted. Platform async calls use /v1/videos/generations/tasks/{taskId}.

Billing succeeds only after a deliverable video URL is available. A temporary polling 5xx is not terminal; treat only status=failed as final failure.

Live Verification

Verified with live requests

Live upstream probe on 2026-07-04: happyhorse-1.0-t2v with 720P, 16:9, duration=3, watermark=false, and seed=123456 reached SUCCEEDED.

The downloaded MP4 was 1280x720, 24 fps, and about 3.04 seconds; successful usage returned duration=3, output_video_duration=3, video_count=1, and SR=720.

StarChasing overseas production regression on 2026-07-06: happyhorse-1.1-t2v was created through /v1/videos/generations, polled to succeeded, and a 5-second video settled as 5,000,000 output tokens.

HappyHorse

happyhorse-1.1-t2v

HappyHorse 1.1 text-to-video

POST /v1/videos/generations

Generates video from text only. Version 1.1 has a lower official 1080P price than 1.0 and is the preferred model for new integrations.

Model price

per video second

720P: $0.14

1080P: $0.18

Alibaba Cloud HappyHorse overseas official USD pricing. The platform maps one video second to 1M output tokens for billing records.Official pricing

Endpoint

POST/v1/videos/generationsPOST/v1/videos/generations/async

Sync and async calls use the same JSON body; for long jobs, create with /async and poll the task API.

ParameterDescription

model

Required. Use the public StarChasing model name on this card; do not send Alibaba Cloud internal aliases.

prompt / content

Prompt text. Prefer a direct prompt string. content text arrays are also accepted and converted into the HappyHorse upstream prompt. Text-to-video, reference-to-video, and video edit require a prompt.

resolution

Optional. Supports 720P or 1080P. The platform also accepts aliases such as 720p, 1080p, hd, fhd, video_resolution, videoResolution, and size. If omitted, StarChasing submits 720P; 480P was rejected in live tests.

seed

Optional integer random seed. The verified valid range is 0 through 2147483647.

watermark

Optional boolean. Set false to disable watermarking.

duration

Optional output duration in seconds. Live probing confirmed it must be between 3 and 15 seconds.

ratio / aspect_ratio

Optional aspect ratio. Verified upstream values are 16:9, 9:16, 4:3, 3:4, 1:1, 5:4, 4:5, 9:21, 21:9.

Each request generates exactly one video. If n is provided and is not 1, the platform rejects the request.

Clients do not need and should not send Alibaba Cloud X-DashScope-Async or Alibaba Cloud API keys; StarChasing adds the upstream async header.

Successful HappyHorse usage is mapped to platform output tokens by video seconds: 1 second = 1,000,000 output tokens; dashboard charges and usage logs follow this rule.

Raw Alibaba Cloud result links are temporary; clients should rely on the video URLs returned by platform task polling, while the platform handles polling and storage.

Official documentation sources

Overview Text-to-video

Minimal text-to-video example

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.1-t2v",
    "prompt": "A red cube slowly rotates on a clean white tabletop, soft studio lighting, simple background.",
    "resolution": "720P",
    "ratio": "16:9",
    "duration": 3,
    "watermark": false,
    "seed": 123456
  }'

# Poll the returned task_id or requestId.
curl https://api.starchasing.ai/v1/videos/generations/task_xxx \
  -H "Authorization: Bearer sk-lg-your-key-here"

HappyHorse

happyhorse-1.1-i2v

HappyHorse 1.1 image-to-video

POST /v1/videos/generations

Generates video from one first-frame image. The image becomes the starting frame, while the prompt controls motion and scene changes.

Model price

per video second

720P: $0.14

1080P: $0.18

Alibaba Cloud HappyHorse overseas official USD pricing. The platform maps one video second to 1M output tokens for billing records.Official pricing

Endpoint

POST/v1/videos/generationsPOST/v1/videos/generations/async

Sync and async calls use the same JSON body; for long jobs, create with /async and poll the task API.

ParameterDescription

model

Required. Use the public StarChasing model name on this card; do not send Alibaba Cloud internal aliases.

prompt / content

resolution

seed

Optional integer random seed. The verified valid range is 0 through 2147483647.

watermark

Optional boolean. Set false to disable watermarking.

first_frame_url / image / image_url

Required for image-to-video: one first-frame image. Public URLs, platform media URLs, and input.media are accepted; the platform converts it to HappyHorse media with type first_frame.

duration

Optional output duration in seconds. Live probing confirmed it must be between 3 and 15 seconds.

Each request generates exactly one video. If n is provided and is not 1, the platform rejects the request.

Clients do not need and should not send Alibaba Cloud X-DashScope-Async or Alibaba Cloud API keys; StarChasing adds the upstream async header.

Successful HappyHorse usage is mapped to platform output tokens by video seconds: 1 second = 1,000,000 output tokens; dashboard charges and usage logs follow this rule.

Raw Alibaba Cloud result links are temporary; clients should rely on the video URLs returned by platform task polling, while the platform handles polling and storage.

Image-to-video does not use ratio as the main control; the aspect ratio follows the first-frame image. Crop the first frame first when a fixed aspect ratio is required.

Official documentation sources

Overview Image-to-video

Minimal image-to-video example

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.1-i2v",
    "prompt": "Turn the product photo into a slow cinematic push-in shot.",
    "first_frame_url": "https://example.com/first-frame.png",
    "resolution": "720P",
    "duration": 5,
    "watermark": false
  }'

HappyHorse

happyhorse-1.1-r2v

HappyHorse 1.1 reference-to-video

POST /v1/videos/generations

Uses multiple reference images to constrain subject, scene, or style, then generates video from the prompt.

Model price

per video second

720P: $0.14

1080P: $0.18

Alibaba Cloud HappyHorse overseas official USD pricing. The platform maps one video second to 1M output tokens for billing records.Official pricing

Endpoint

POST/v1/videos/generationsPOST/v1/videos/generations/async

Sync and async calls use the same JSON body; for long jobs, create with /async and poll the task API.

ParameterDescription

model

Required. Use the public StarChasing model name on this card; do not send Alibaba Cloud internal aliases.

prompt / content

resolution

seed

Optional integer random seed. The verified valid range is 0 through 2147483647.

watermark

Optional boolean. Set false to disable watermarking.

reference_images / images

For reference-to-video, pass an array of reference images. Live probing confirmed a maximum of 9 images; the platform converts them to HappyHorse media with type reference_image.

duration

Optional output duration in seconds. Live probing confirmed it must be between 3 and 15 seconds.

ratio / aspect_ratio

Optional aspect ratio. Verified upstream values are 16:9, 9:16, 4:3, 3:4, 1:1, 5:4, 4:5, 9:21, 21:9.

Each request generates exactly one video. If n is provided and is not 1, the platform rejects the request.

Clients do not need and should not send Alibaba Cloud X-DashScope-Async or Alibaba Cloud API keys; StarChasing adds the upstream async header.

Successful HappyHorse usage is mapped to platform output tokens by video seconds: 1 second = 1,000,000 output tokens; dashboard charges and usage logs follow this rule.

Raw Alibaba Cloud result links are temporary; clients should rely on the video URLs returned by platform task polling, while the platform handles polling and storage.

Official documentation sources

Overview Reference-to-video

Minimal reference-to-video example

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.1-r2v",
    "prompt": "Use the reference images to keep the same character and create a short runway walk video.",
    "reference_images": [
      "https://example.com/reference-1.png",
      "https://example.com/reference-2.png"
    ],
    "resolution": "1080P",
    "ratio": "9:16",
    "duration": 5,
    "watermark": false
  }'

HappyHorse

happyhorse-1.0-t2v

HappyHorse 1.0 text-to-video

POST /v1/videos/generations

Version 1.0 text-to-video is kept for workloads that need the older model behavior. For new work, evaluate 1.1 first.

Model price

per video second

720P: $0.14

1080P: $0.24

Alibaba Cloud HappyHorse overseas official USD pricing. The platform maps one video second to 1M output tokens for billing records.Official pricing

Endpoint

POST/v1/videos/generationsPOST/v1/videos/generations/async

Sync and async calls use the same JSON body; for long jobs, create with /async and poll the task API.

ParameterDescription

model

Required. Use the public StarChasing model name on this card; do not send Alibaba Cloud internal aliases.

prompt / content

resolution

seed

Optional integer random seed. The verified valid range is 0 through 2147483647.

watermark

Optional boolean. Set false to disable watermarking.

duration

Optional output duration in seconds. Live probing confirmed it must be between 3 and 15 seconds.

ratio / aspect_ratio

Optional aspect ratio. Verified upstream values are 16:9, 9:16, 4:3, 3:4, 1:1, 5:4, 4:5, 9:21, 21:9.

Each request generates exactly one video. If n is provided and is not 1, the platform rejects the request.

Clients do not need and should not send Alibaba Cloud X-DashScope-Async or Alibaba Cloud API keys; StarChasing adds the upstream async header.

Successful HappyHorse usage is mapped to platform output tokens by video seconds: 1 second = 1,000,000 output tokens; dashboard charges and usage logs follow this rule.

Raw Alibaba Cloud result links are temporary; clients should rely on the video URLs returned by platform task polling, while the platform handles polling and storage.

Official documentation sources

Overview Text-to-video

Minimal text-to-video example

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.0-t2v",
    "prompt": "A red cube slowly rotates on a clean white tabletop, soft studio lighting, simple background.",
    "resolution": "720P",
    "ratio": "16:9",
    "duration": 3,
    "watermark": false,
    "seed": 123456
  }'

# Poll the returned task_id or requestId.
curl https://api.starchasing.ai/v1/videos/generations/task_xxx \
  -H "Authorization: Bearer sk-lg-your-key-here"

HappyHorse

happyhorse-1.0-i2v

HappyHorse 1.0 image-to-video

POST /v1/videos/generations

Version 1.0 image-to-video uses one first-frame image; the client request shape is the same as 1.1 i2v.

Model price

per video second

720P: $0.14

1080P: $0.24

Alibaba Cloud HappyHorse overseas official USD pricing. The platform maps one video second to 1M output tokens for billing records.Official pricing

Endpoint

POST/v1/videos/generationsPOST/v1/videos/generations/async

Sync and async calls use the same JSON body; for long jobs, create with /async and poll the task API.

ParameterDescription

model

Required. Use the public StarChasing model name on this card; do not send Alibaba Cloud internal aliases.

prompt / content

resolution

seed

Optional integer random seed. The verified valid range is 0 through 2147483647.

watermark

Optional boolean. Set false to disable watermarking.

first_frame_url / image / image_url

Required for image-to-video: one first-frame image. Public URLs, platform media URLs, and input.media are accepted; the platform converts it to HappyHorse media with type first_frame.

duration

Optional output duration in seconds. Live probing confirmed it must be between 3 and 15 seconds.

Each request generates exactly one video. If n is provided and is not 1, the platform rejects the request.

Clients do not need and should not send Alibaba Cloud X-DashScope-Async or Alibaba Cloud API keys; StarChasing adds the upstream async header.

Successful HappyHorse usage is mapped to platform output tokens by video seconds: 1 second = 1,000,000 output tokens; dashboard charges and usage logs follow this rule.

Raw Alibaba Cloud result links are temporary; clients should rely on the video URLs returned by platform task polling, while the platform handles polling and storage.

Image-to-video does not use ratio as the main control; the aspect ratio follows the first-frame image. Crop the first frame first when a fixed aspect ratio is required.

Official documentation sources

Overview Image-to-video

Minimal image-to-video example

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.0-i2v",
    "prompt": "Turn the product photo into a slow cinematic push-in shot.",
    "first_frame_url": "https://example.com/first-frame.png",
    "resolution": "720P",
    "duration": 5,
    "watermark": false
  }'

HappyHorse

happyhorse-1.0-r2v

HappyHorse 1.0 reference-to-video

POST /v1/videos/generations

Version 1.0 reference-to-video uses 1-9 reference images; the client request shape is the same as 1.1 r2v.

Model price

per video second

720P: $0.14

1080P: $0.24

Alibaba Cloud HappyHorse overseas official USD pricing. The platform maps one video second to 1M output tokens for billing records.Official pricing

Endpoint

POST/v1/videos/generationsPOST/v1/videos/generations/async

Sync and async calls use the same JSON body; for long jobs, create with /async and poll the task API.

ParameterDescription

model

Required. Use the public StarChasing model name on this card; do not send Alibaba Cloud internal aliases.

prompt / content

resolution

seed

Optional integer random seed. The verified valid range is 0 through 2147483647.

watermark

Optional boolean. Set false to disable watermarking.

reference_images / images

For reference-to-video, pass an array of reference images. Live probing confirmed a maximum of 9 images; the platform converts them to HappyHorse media with type reference_image.

duration

Optional output duration in seconds. Live probing confirmed it must be between 3 and 15 seconds.

ratio / aspect_ratio

Optional aspect ratio. Verified upstream values are 16:9, 9:16, 4:3, 3:4, 1:1, 5:4, 4:5, 9:21, 21:9.

Each request generates exactly one video. If n is provided and is not 1, the platform rejects the request.

Clients do not need and should not send Alibaba Cloud X-DashScope-Async or Alibaba Cloud API keys; StarChasing adds the upstream async header.

Successful HappyHorse usage is mapped to platform output tokens by video seconds: 1 second = 1,000,000 output tokens; dashboard charges and usage logs follow this rule.

Raw Alibaba Cloud result links are temporary; clients should rely on the video URLs returned by platform task polling, while the platform handles polling and storage.

Official documentation sources

Overview Reference-to-video

Minimal reference-to-video example

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.0-r2v",
    "prompt": "Use the reference images to keep the same character and create a short runway walk video.",
    "reference_images": [
      "https://example.com/reference-1.png",
      "https://example.com/reference-2.png"
    ],
    "resolution": "1080P",
    "ratio": "9:16",
    "duration": 5,
    "watermark": false
  }'

HappyHorse

happyhorse-1.0-video-edit

HappyHorse 1.0 video edit

POST /v1/videos/generations

Edits a source video using a prompt and optional reference images. StarChasing currently verifies video plus image references only and does not expose Alibaba Cloud audio controls.

Model price

per video second

720P: $0.14

1080P: $0.24

Alibaba Cloud HappyHorse overseas official USD pricing. The platform maps one video second to 1M output tokens for billing records.Official pricing

Endpoint

POST/v1/videos/generationsPOST/v1/videos/generations/async

Sync and async calls use the same JSON body; for long jobs, create with /async and poll the task API.

ParameterDescription

model

Required. Use the public StarChasing model name on this card; do not send Alibaba Cloud internal aliases.

prompt / content

resolution

seed

Optional integer random seed. The verified valid range is 0 through 2147483647.

watermark

Optional boolean. Set false to disable watermarking.

video_url / video

Required for video edit: one source video. The platform converts it to HappyHorse media with type video; live probing returned a terminal failure when it was missing.

reference_images / images

Optional reference images for video edit, used for style or subject guidance. The public platform currently exposes image references only and does not expose Alibaba Cloud audio_setting.

Each request generates exactly one video. If n is provided and is not 1, the platform rejects the request.

Clients do not need and should not send Alibaba Cloud X-DashScope-Async or Alibaba Cloud API keys; StarChasing adds the upstream async header.

Successful HappyHorse usage is mapped to platform output tokens by video seconds: 1 second = 1,000,000 output tokens; dashboard charges and usage logs follow this rule.

Raw Alibaba Cloud result links are temporary; clients should rely on the video URLs returned by platform task polling, while the platform handles polling and storage.

Do not rely on duration to change the source video length for video edit; per official docs, only the first 15 seconds are processed when the input video is longer than 15 seconds.

Official documentation sources

Overview Video edit

Minimal video edit example

curl https://api.starchasing.ai/v1/videos/generations \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.0-video-edit",
    "prompt": "Replace the background with a bright studio scene while keeping the main subject motion natural.",
    "video_url": "https://example.com/source-video.mp4",
    "reference_images": ["https://example.com/style-reference.png"],
    "resolution": "720P",
    "watermark": false
  }'

Dashboard Workflow

Dashboard HappyHorse video generation

Open Video generation in the StarChasing dashboard and choose a happyhorse-* model. Do not enter Alibaba Cloud task_id, X-DashScope-Async, or asset:// in the web UI.

t2v uses no media; i2v uploads one first-frame image; r2v uploads 1-9 reference images; video-edit uploads one video and can add reference images.

After submission, follow the platform task status. In-progress work stays running; after success, the dashboard shows the platform-stored video and usage logs show requestId, video seconds, and charges.

Generation

Generation API modes

Synchronous generation endpoints keep the existing OpenAI-compatible behavior: the request waits for upstream directly, does not create a public async task, and is not affected by the async queue. Current async video sync routes, including Seedance and HappyHorse, do not require clients to parse provider-specific task envelopes; clients should rely on top-level id, task_id, status, and video_urls. When an upstream task id is available, the platform returns status=running first and settles only after deliverable video media is available. For known in-progress async video tasks, a temporary upstream polling 5xx stays status=running and must not be treated as terminal failure; during submission, an async video upstream HTTP 200 without either a task id or a video URL is recorded as UPSTREAM_VIDEO_RESULT_MISSING for the current route, then the request continues through fallback.

Async generation endpoints append /async to the original path: creation returns HTTP 202, then the client polls /tasks/{taskId} using id or status_url.

Polling only reads the platform task state and does not call upstream directly. Queueing, upstream submission, video polling, media storage, and final settlement are handled by workers.

Task statuses include queued, submitted, running, storing_media, succeeded, failed, expired, and cancelled; internal capacity wait reasons are not exposed to users.

Async Example

Create and poll a task

curl https://api.starchasing.ai/v1/images/generations/async \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-5-0-lite",
    "prompt": "A clean product photo of a ceramic coffee cup",
    "size": "3K",
    "n": 1
  }'

# Use the returned id, for example task_xxx.
curl https://api.starchasing.ai/v1/images/generations/tasks/task_xxx \
  -H "Authorization: Bearer sk-lg-your-key-here"

Successful creation response

{
  "id": "task_xxx",
  "object": "generation.task",
  "kind": "image_generation",
  "status": "queued",
  "status_url": "/v1/images/generations/tasks/task_xxx",
  "request_id": null,
  "llmgate_image": {
    "requested": {
      "model": "doubao-seedream-5-0-lite",
      "size": "3K",
      "n": 1
    },
    "standard": {
      "model": "doubao-seedream-5-0-lite",
      "size": "3k",
      "image_size_tier": "3K",
      "n": 1
    }
  },
  "result": null,
  "error": null
}

Dashboard

Dashboard Guides

This section only covers dashboard workflows. API integration details stay in the model cards and API mode section above.

Dashboard Workflow

First/last frame video generation

This section is for dashboard users; they do not need to understand upstream payloads. The platform turns uploaded images into the Seedance first-frame and last-frame roles.

Open Video generation in the dashboard and choose doubao-seedance-2-0 or doubao-seedance-2-0-fast.

Switch the reference mode to First/last frame; this mode only accepts images, not reference video or audio.

Upload the first image as the first frame; upload a second image when a last frame is needed. A task cannot use a last frame without a first frame.

Enter the prompt, duration, aspect ratio, and resolution, then submit. The platform handles image upload, route-specific asset libraries, and task status; users do not enter asset://.

Minimum Setup

Minimum API setup

Use Authorization: Bearer sk-lg-your-key-here in request headers.

Set base_url in the Python OpenAI SDK and baseURL in the Node.js OpenAI SDK.

Send chat completion requests to /chat/completions; the full URL is joined with base_url.

Use the platform model name, for example glm-5.

After the call, verify requestId, tokens, charges, and product channel in dashboard usage logs.

Python example: change two lines

API endpointhttps://api.starchasing.ai/v1

from openai import OpenAI

client = OpenAI(
    api_key="sk-lg-your-key-here",
    base_url="https://api.starchasing.ai/v1",
)

response = client.chat.completions.create(
    model="glm-5",
    messages=[{"role": "user", "content": "Hello!"}],
)

print(response.choices[0].message.content)

Video Input

Chat video input

`/v1/chat/completions` supports controlled video understanding. The public standard content part uses `type=video`, `uri` or `data`, and `mime_type`; the model must expose `capabilities.video_input` from `/v1/models`. If the selected route cannot handle video input, the platform falls back to other available candidates.

curl https://api.starchasing.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-lg-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-2.5-pro",
    "messages": [
      {
        "role": "user",
        "content": [
          {"type": "text", "text": "Describe this video in chronological order."},
          {"type": "video", "uri": "https://example.com/video.mp4", "mime_type": "video/mp4"}
        ]
      }
    ]
  }'

API docs and model pricing.

API integration flow

管理员创建账号并充值

Create API Key

Replace base_url

Send a test request

Common supported API endpoints

Text model calls and pricing

glm-5

glm-5.1

glm-5.2

deepseek-v3.2

deepseek-v4-flash

deepseek-v4-pro

Image generation models

Doubao Seedream 5.0 Lite

Doubao Seedream 4.5

Video generation models

Doubao Seedance 2.0

Doubao Seedance 2.0 Fast

HappyHorse video generation manual

Create a task, then poll the result

Verified with live requests

happyhorse-1.1-t2v

happyhorse-1.1-i2v

happyhorse-1.1-r2v

happyhorse-1.0-t2v

happyhorse-1.0-i2v

happyhorse-1.0-r2v

happyhorse-1.0-video-edit

Dashboard HappyHorse video generation

Generation API modes

Create and poll a task

Dashboard Guides

First/last frame video generation

Minimum API setup

Chat video input