Creator Tools

Transcript Maker

YouTube Audio-to-Subtitle Workspace

Flask and yt-dlp speech-to-text transcript alignment, SRT timing editor, and Sinhala translation suite.

Overview

Transcript Maker is an audio and video transcription workspace engineered for content creators, subtitle editors, and media translators. It ingests any YouTube video URL or local audio file, extracts high-precision timecoded transcripts, and provides a dual-pane editor for text correction, word-level timing adjustments, and phonetic Sinhala translation with instant SRT and VTT export.

Design decisions

The choices that shaped the system, and why.

1
Why Local File Caching for Transcripts?
Querying YouTube API repeatedly triggers strict rate limits and IP blocks. Implemented deterministic SHA-256 caching of fetched subtitle tracks and audio streams to ensure instant repeat edits.
2
Interactive Click-to-Seek Audio Alignment
Integrated HTML5 media player event listeners with rendered subtitle spans so clicking any translated word seeks audio to that exact millisecond.

Architecture

Flask & yt-dlp Backend Service

Asynchronously fetches video streams, captions, and audio metadata with localized caching.

Precise Audio-Text Alignment Engine

Synchronizes interactive transcript spans with an HTML5 media player for click-to-seek playback.

Bilingual Translation Subsystem

Translates English dialogue into conversational Sinhala while preserving timing markers.

Export Pipeline

Formats clean SRT, WebVTT, and plain text transcripts ready for YouTube Studio or Premiere Pro.

DevOps & deployment

Containerized with Docker; deployable on GCP Cloud Run or AWS App Runner.

How it ships · Docker / Gunicorn

1docker compose up --build
2Gunicorn serves Flask on :5050
3Transcript cache persisted on a volume

How I ship my projects

Features

Sub-second transcript extraction for any public YouTube video.
Interactive waveform / audio player with click-to-jump subtitle sync.
Editable translation columns with auto-save and undo/redo.
Zero cloud lock-in: fully containerized and runnable offline.

Screenshots

Fig 1 · Subtitle editor with Sinhala translation

Fig 2 · Transcript workspace — paste a YouTube link to start

Related writing

Deep dive · 4 min