Tutorial · Python · ~10 min

How to Remove Hardcoded Subtitles from Video in Python

Hardcoded subtitles are rendered into the pixels, so they can't be toggled off. This tutorial erases them with frame-aware AI inpainting — no masking, no cropping, no manual rotoscoping.

PythonSubtitle removalInpainting

Erase subtitles with AI

Before you start

Prerequisites

A free mlslabs account and API key
Python 3.9+
pip install mlslabs
No footage needed — use our cloud sample (short-drama-cn.mp4)

Step 1 — Install the SDK

Bash

pip install mlslabs

Step 2 — Set your API key

Python

import mlslabs

client = mlslabs.Client("YOUR_API_KEY")

Step 3 — Submit an erasure job

mode="subtitle" targets caption regions automatically. If detection misses stylized text, pass explicit region boxes instead.

Python

job = client.subtitle_eraser.submit(
    input_url="s3://bucket/short-drama-cn.mp4",
    output_url="s3://bucket/short-drama-clean.mp4",
    mode="subtitle",   # subtitle | text | region boxes
)
print(job.job_id)

Step 4 — Poll until complete

Python

job.wait()
print(job.status)  # done | failed
# failed jobs auto-refund credits — check job.error and retry

Step 5 — Download and verify

No residual text, blur or color blocks where captions were
No flicker across scene cuts
Resolution matches the source exactly

Troubleshooting

Symptom	Cause	Fix
No text detected	Caption style outside training	Use mode="text" or region boxes
Partial erasure	Tight detection box	Box 30–50% larger than the text
Job failed	Transient error	Check job.error; credits were refunded — retry

Run this tutorial in Postman — the erasure request is prefilled with the sample episode.

Keep exploring

Related guides

Subtitle Eraser API

Remove hardcoded subtitles cleanly for localization and short-drama export.

Learn more

Subtitle Extractor API

Extract on-screen text and spoken dialogue with frame-accurate timestamps.

Learn more

SDR2HDR API

Convert SDR video to HDR with restored brightness, gamut and detail.

Learn more