---

CapScript Pro, GUI Based YouTube Transcript Search Tool

CapScript Pro

CapScript Pro is a Windows desktop application for searching YouTube captions at scale, then turning those findings into usable media outputs.

This repository includes:

• A modern Qt/C++ desktop app.
• An embedded Python engine for yt-dlp transcript fetching and search logic.
• A standalone Python CLI for headless/automation workflows.

Table of Contents

• Overview
• Desktop App Features
• How the Embedded Python Engine Works
• Tech Stack and Dependencies
• Project Structure
• Desktop App Setup
• Build and Run (Qt App)
• CLI Guide
• CLI Setup and Requirements
• CLI Usage
• CLI Examples
• Troubleshooting
• License

Overview

CapScript Pro solves a practical workflow:

1. Find subtitle matches for a keyword across videos.
2. Review exact timestamps inside the built-in viewer.
3. Extract clips from the matched moments.
4. Render clips into a final output video.
5. Build reusable video ID lists for future runs.

The UI is native Qt (C++), while core YouTube logic is powered by Python.

v2.5 — No more YouTube Data API. The app no longer uses the YouTube Data API at all. It was too restrictive, had an embarrassingly low quota, and broke constantly — sometimes for no apparent reason. Everything now runs through yt-dlp directly, which is faster, more reliable, and doesn't require you to register an API key, manage credentials, or hit an arbitrary daily limit. Good riddance.

Desktop App Features

• Search Page: Handles flexible search queries for videos or channels. Includes advanced network settings (proxy, cookies), language selection, and encrypted credential persistence.
• Viewer Page: Displays structured transcripts with clickable timestamps and supports in-app video playback. Allows users to export or import transcript files.
• Clip Downloader: Parses transcripts to identify specific clips for download. Supports various formats (mp4, mkv, webm, mp3), quality controls, and uses worker threads for efficient background processing.
• Renderer Page: Merges downloaded clips into a single final video using FFmpeg. Auto-detects formats and utilizes stream copying for fast, lossless concatenation.
• List Creator: Fetches channel videos based on date ranges or keywords. Provides a selectable list with thumbnail previews and options to export video IDs.
• System & Support: Includes an integrated update checker that stages updates automatically.

How the Embedded Python Engine Works

The desktop app starts a bundled Python runtime and imports capscript_engine.py through a C++ bridge.

High-level flow:

1. The app resolves Python runtime paths near the executable.
2. PythonBridge initializes Python with controlled module search paths.
3. C++ invokes Python functions for:
   • Proxy settings storage/loading.
   • Channel resolution and video lookups.
   • Transcript search execution.
4. Search progress is sent back into C++ via callback trampolines and displayed in UI/CLI.

This split keeps the UI native and fast while preserving Python ecosystem advantages for YouTube-related operations.

Tech Stack and Dependencies

Desktop Application

• C++17
• Qt 6 modules:
  • Core, Gui, Widgets, Network, Concurrent, Svg
  • Optional: Multimedia, MultimediaWidgets
  • Optional: Quick, Qml, QuickWidgets, WebView, WebViewQuick
• Python 3 interpreter + development libs for embedding
• Windows WebView2 runtime/SDK (optional but recommended for in-app player)

Python Engine and CLI

From python/requirements.txt:

• yt-dlp
• cryptography
• rich

External Media Tools

Required for clip download/render features:

• yt-dlp
• ffmpeg
• ffprobe (normally included with ffmpeg)

The app looks for tools in:

• app/bin/
• app root
• system PATH

Optional Service Component

Cloudflare Worker (TypeScript) for feedback delivery and basic rate limiting.

Project Structure

• src/: Qt C++ application and workers
• python/: capscript_engine.py + cli.py + Python requirements
• scripts/: utility scripts (including embedded runtime bundling)
• qml/: QML resources for web player integrations
• assets/: app fonts, icons, resources
• cloudflare-worker/: optional feedback backend
• docs/: web documentation assets
• third_party/webview2/: WebView2 SDK assets

Desktop App Setup

Prerequisites

• Windows 10/11
• CMake 3.21+
• Qt 6 (matching your compiler toolchain)
• Python 3 with headers/libs discoverable by CMake
• Optional but recommended: WebView2 runtime installed
• Optional for clip/render workflow: yt-dlp and ffmpeg available

Optional: Bundle Python Runtime for Distribution

Use the provided script to package embeddable Python + pip dependencies into a target python folder:

python scripts/bundle_python.py 3.11.9 .\python .\python\requirements.txt

Build and Run (Qt App)

Example CMake flow (Visual Studio generator):

cmake -S . -B build -G "Visual Studio 17 2022" -A x64
cmake --build build --config Release

Run the desktop app from build output after ensuring python/ runtime assets are present near the executable.

UI

CLI Guide

The repository includes a full command-line interface in python/cli.py for non-GUI usage.

Use CLI when you need:

• Automation (batch jobs, scripting, CI-like tasks).
• Remote/headless execution.
• Fast one-off searches without opening the desktop app.

CLI Setup and Requirements

1. Python Version

Recommended: Python 3.10+.

2. Install Dependencies

cd python
python -m pip install -r requirements.txt

3. Authentication Model

No API key is required. CapScript Pro runs entirely through yt-dlp.

For authenticated requests (age-restricted content, private playlists, etc.) pass cookies directly from your browser — no manual export needed:

--cookies-from-browser chrome|firefox|edge|brave

Legacy API-key CLI flags are accepted as no-ops for compatibility with older scripts.

4. Optional Network Setup

• Cookies file (Netscape/Mozilla format): --cookies path\to\cookies.txt
• Cookies from browser (new): --cookies-from-browser chrome|firefox|edge|brave
• Proxy modes:
  • --proxy-type webshare with username/password
  • --proxy-type generic with proxy URL
• Persist proxy settings with --save-proxy
• Clear saved proxy with --clear-proxy

CLI Usage

Main Search Command

python cli.py --search-type channel --channel "@mkbhd" --keyword "sponsors" --max-results 20

Required Inputs

• --search-type channel or video
• --keyword "your search term"
• For channel mode: --channel
• For video mode: --video-ids

Useful Options

• --language en
• --output-dir transcripts
• --cookies FILE
• --cookies-from-browser chrome|firefox|edge|brave
• --proxy-type, --proxy-username, --proxy-password, --proxy-url
• --save-proxy, --clear-proxy

CLI Examples

1. Search a channel by handle

python cli.py --search-type channel --channel "@mkbhd" --keyword "sponsors" --max-results 20

2. Search explicit video URLs/IDs

python cli.py --search-type video --video-ids "https://youtu.be/dQw4w9WgXcQ,abc123XYZ00" --keyword "never gonna"

3. Use browser cookies

python cli.py --search-type channel --channel "@mkbhd" --keyword "AI" --cookies-from-browser chrome

4. Use Webshare proxy + cookies

python cli.py --search-type channel --channel "UCxxxxxx" --keyword "AI" --proxy-type webshare --proxy-username USER --proxy-password PASS --cookies cookies.txt

5. Save generic proxy once

python cli.py --proxy-type generic --proxy-url "http://1.2.3.4:8080" --save-proxy
python cli.py --search-type video --video-ids abc123 --keyword test

Troubleshooting

• Python engine fails to initialize in app: Ensure python/ exists near the executable and contains runtime artifacts (Lib, python3xx zip, required modules).
• No transcript results: Try --cookies-from-browser or a cookies.txt file, and verify the language code is available for each video.
• Clip download/render not working: Confirm yt-dlp and ffmpeg are installed and discoverable. Check app logs for tool path and process errors.
• Rate limiting or access errors: Use --cookies-from-browser to pass your browser session. Reduce parallel activity and retry later.

Bugs

Please report bugs by opening an issue on GitHub.

License

This project is licensed under the MIT License + Commons Clause v1.0.

You are free to use, modify, and distribute this software for free, but you may not sell it or offer it as a paid/SaaS service.

See the LICENSE file for the full text.