Skip to main content
POST /v1/audio/transcriptions Send multipart containing model and the file bytes. Preserve the correct filename extension; a local path string is not a file upload. response_format: json returns text; verbose_json can include duration/segments. Plain-text or subtitle formats return text directly. language describes the input language. Confirm timestamps, diarization, and streaming separately; this reference documents the non-streaming baseline. Integration guide and runnable example 简体中文

Authorizations

​
Authorization
string
header
required

模型调用 API Key，不是控制台登录凭据 / Model API key, not console login credentials

Body

multipart/form-data
​
model
string
required

当前密钥可用且支持此接口的模型 ID / Model ID available to the key and supporting this endpoint

Example:

"YOUR_MODEL_ID"

​
file
file
required

上传音频文件，保留正确扩展名；不是路径字符串或 URL / Uploaded audio file with correct extension, not a path string or URL

​
response_format
string

示例使用 json；text、srt、vtt、verbose_json 等需模型支持 / Examples use json; text, srt, vtt, verbose_json and other formats require model support

​
prompt
string

可选上下文提示，不是翻译目标语言 / Optional context hint, not the target translation language

​
temperature
number

采样温度，需模型支持 / Sampling temperature when supported

​
language
string

输入语言提示，例如 zh、en；模型决定格式 / Input-language hint such as zh or en; model determines format

​
timestamp_granularities[]
string[]

word 或 segment，通常配合 verbose_json；需模型支持 / word or segment, usually with verbose_json; requires support

Response

由 response_format 决定；示例为 json / Determined by response_format; example uses json

​
text
string
required

转写或翻译文本，静音可能为空 / Transcribed or translated text; silence may yield empty text