API 调用文档

聊天 / Codex

使用 OpenAI 兼容的 Chat Completions 接口。把 model 换成下面模型表里的模型名即可。

curl https://ai.yuyapi.cn/v1/chat/completions \
  -H "Authorization: Bearer sk-你的令牌" \
  -H "Content-Type: application/json" \
  -d '{"model":"模型名","messages":[{"role":"user","content":"你好"}]}'

图片生成

gpt-image-2 是异步图片任务：提交成功时通常只返回 task_id，不会立即返回图片 URL。下游软件不要从提交接口直接读取 data[0].url；必须保存 task_id，再轮询任务结果。

curl https://ai.yuyapi.cn/v1/images/generations \
  -H "Authorization: Bearer sk-你的令牌" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-image-2","prompt":"一张城市夜景","n":1,"size":"1024x1024","quality":"auto","response_format":"url"}'

# 提交成功示例
{"id":"task_xxx","task_id":"task_xxx","object":"image.task","status":"queued","model":"gpt-image-2"}

curl https://ai.yuyapi.cn/v1/images/tasks/task_xxx \
  -H "Authorization: Bearer sk-你的令牌"

# 轮询成功后再取图
{"id":"task_xxx","task_id":"task_xxx","status":"succeeded","progress":100,"result":{"data":[{"url":"https://example.com/image.png"}]}}

成功状态为 succeeded 或 completed。推荐读取 result.data[0].url，同时兼容 data[0].url、result.url、url。处理中状态 queued / processing / running 需要继续轮询。

图生图必须使用 multipart/form-data，多个参考图重复传 image[]。

curl https://ai.yuyapi.cn/v1/images/edits \
  -H "Authorization: Bearer sk-你的令牌" \
  -F "model=gpt-image-2" \
  -F "prompt=参考这张图生成新图" \
  -F "image=@/path/to/input.jpg"

T香蕉2、T香蕉pro 使用 Gemini 原生异步生图接口：POST /v1beta/models/{model}:generateContent，提交成功后继续轮询 GET /v1/images/tasks/{task_id}。多图参考时把图片 base64 放进 parts[].inlineData.data。

curl https://ai.yuyapi.cn/v1beta/models/T香蕉2:generateContent \
  -H "Authorization: Bearer sk-你的令牌" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"role":"user","parts":[{"text":"生成一张写实摄影风格的橙色跑车，城市夜景"}]}],"generationConfig":{"responseModalities":["TEXT","IMAGE"],"imageConfig":{"aspectRatio":"16:9","imageSize":"2K"}}}'

视频生成

Sora-v3 使用 POST /v1/video/submit/generate 和 GET /v1/video/fetch/{task_id}。支持文生视频、单图图生视频和首尾帧视频。

curl https://ai.yuyapi.cn/v1/video/submit/generate \
  -H "Authorization: Bearer sk-你的令牌" \
  -H "Content-Type: application/json" \
  -d '{"model":"sora-v3-fast","prompt":"一段电影感视频","duration":8,"metadata":{"modeType":"text2video","ratio":"16:9","enableSound":"on"}}'

curl https://ai.yuyapi.cn/v1/video/fetch/task_xxx \
  -H "Authorization: Bearer sk-你的令牌"

Grok、Omni、Veo、KY、XH、LD 使用 POST /v1/videos 和 GET /v1/videos/{task_id}。参考图统一传 images 数组；需要视频/音频参考的新模型可传 videos、audios 数组；Veo 不需要 seconds；Omni Edit 需要 video_url。

curl https://ai.yuyapi.cn/v1/videos \
  -H "Authorization: Bearer sk-你的令牌" \
  -H "Content-Type: application/json" \
  -d '{"model":"omni-flash-components-1080p","prompt":"make the subject walk forward naturally","seconds":"8","images":["https://example.com/input.jpg"],"aspect_ratio":"9:16","n":1}'

curl https://ai.yuyapi.cn/v1/videos/task_xxx \
  -H "Authorization: Bearer sk-你的令牌"