Extract Chinese Subtitles from Video

Why this matters

Chinese characters are dense and visually complex — generic OCR models mangle them. A zh-specialized model handles simplified and traditional forms, punctuation, and vertical layouts.

For short dramas, the burned-in zh subtitles are the source of truth for every target language; extraction quality sets the ceiling for MT quality.

How it works

Set lang=zh on the job. The API OCRs with a zh model, clusters by timing, and returns SRT with confidence scores.

Submit a job with an input URL (S3, GCS or HTTPS), poll the job URL, and download the rendered output. No GPU, no queues, no ffmpeg builds to babysit.

Code

import requests, time

API = "https://api.mlslabs.io/v1/subtitle-extractor/jobs"
headers = {"X-API-Key": "YOUR_API_KEY"}
payload = {"input": "s3://bucket/video.mp4", "output_format": "srt", "lang": "zh"}

resp = requests.post(API, json=payload, headers=headers)
job = resp.json()
while job["status"] not in ("succeeded", "failed"):
    time.sleep(3)
    job = requests.get(job["url"], headers=headers).json()

print("Output:", job["output_url"])

Pro tips

• Feed the SRT to your MT pipeline directly — it keeps line boundaries and timing.
• Flag low-confidence lines for a human QC pass; they are usually 5% or less.
• Works for both horizontal and vertical (9:16) subtitle layouts.

Try zh extraction in Postman.