Python 客户端

Python 客户端封装了公开上传接口和任务接口，并提供同步与异步客户端。

安装

pip install video-to-text-python

初始化

import os

from video_to_text import VideoToTextClient

client = VideoToTextClient(
    api_key=os.environ["VTT_API_KEY"],
)

转录文件

from pathlib import Path

task = client.transcriptions.transcribe_file(
    Path("meeting.mp4"),
    language="Auto",
    timestamp_mode="CHUNK",
    transcription_mode="balanced",
    timeout=600,
)

print(task.full_text)

当准确性比学分成本更重要时，可以使用 precision。

手动上传并创建任务

upload = client.uploads.create({
    "filename": "meeting.mp4",
    "mimetype": "video/mp4",
})

content = Path("meeting.mp4").read_bytes()
client.uploads.put(upload.upload_url, content, mimetype="video/mp4")

asset = client.uploads.complete(upload.upload_id, {
    "fileKey": upload.file_key,
    "fileUrl": upload.file_url,
    "filename": "meeting.mp4",
    "mimetype": "video/mp4",
    "fileSize": len(content),
})

created = client.tasks.create({
    "assetId": asset.asset_id,
    "language": "Auto",
    "timestampMode": "CHUNK",
    "transcriptionMode": "balanced",
})

result = client.tasks.wait(created.transcript_id)
print(result.full_text)

异步客户端

from video_to_text import AsyncVideoToTextClient

async with AsyncVideoToTextClient(
    api_key=os.environ["VTT_API_KEY"],
) as client:
    task = await client.transcriptions.transcribe_file(
        Path("meeting.mp4"),
        transcription_mode="balanced",
    )
    print(task.full_text)

任务结果

tasks.wait() 会在任务进入 SUCCEEDED 时返回。如果任务进入 FAILED 或 CANCELED，客户端会抛出 APIError，并优先使用 show_error 作为错误详情。

任务详情包含 full_text、chunks、words、source_duration_ms、result_language、transcription_mode 和 billed_credits。

错误处理

from video_to_text import APIError, RateLimitError, UploadError

try:
    task = client.transcriptions.transcribe_file(
        Path("meeting.mp4"),
        transcription_mode="balanced",
    )
except RateLimitError:
    print("Please retry later.")
except UploadError:
    print("The media upload did not complete.")
except APIError as error:
    print(error.error_code, error.problem.detail)

资源

源代码、问题反馈和贡献请查看 video-to-text-python GitHub 仓库。最新版本可从 PyPI 上的 video-to-text-python 安装。