YTArchiver

Download, organize, transcribe, search, compress, and browse entire channels

(patches will pause for a bit, will continue to report issues until I can get them patched)

Features

Downloading & Syncing

• Channel subscriptions — add channels to a sub list with per-channel settings
• Selective archiving — set a start date, grab the full history, or only pull new uploads
• Auto-sync — set a recurring interval (1hr, 3hr, 6hr, etc.) to automatically check all subbed channels for new videos
• Duration filters — exclude Shorts, livestreams, or long-form videos by setting min/max duration limits
• Resolution control — per-channel resolution settings (144p–1080p, or "best")
• Redownload — change a channel's resolution and retroactively redownload all videos at the new setting
• Resumable downloads — ID caching, so interrupted syncs pick back up without re-scanning
• Cookie support — uses Firefox cookies for age-restricted or member content, and helps avoid IP rate-limiting

Organization

• Folder sorting — sort videos into \YYYY\ or \YYYY\MM\ folders, configurable per channel
• Reorganize tool — re-sort existing downloads into a new org structure at any time
• Date Fix — retroactively set file dates to the original YouTube upload date (fuzzy title matching, useful for migrating from other tools)

Transcription

• Auto-captions first — pulls YouTube's built-in captions when available, with punctuation model cleanup
• Whisper GPU fallback — runs Whisper locally on GPU for videos without captions; model selectable per channel
• Auto-transcribe — per-channel toggle to automatically transcribe new videos after each sync
• Transcript output — clean .txt files with an option to follow the channel's folder org or combine into a single file per channel
• Hidden JSONL sidecars — per-word timestamps and video IDs stored alongside readable transcripts

Browse Tab

• Searchable transcript database — full-text search across all transcribed channels
• Embedded video player — HTML5 video with a synced, scrolling transcript alongside
• Click-to-seek — click any word in the transcript to jump the video to that moment
• Word frequency analysis — frequency graphs (Year / Month / Week buckets) and word cloud visualizations
• Recent downloads — list or thumbnail-grid view of recently-archived videos, filterable

Compression

• AV1 NVENC encoding — compress archived videos using AV1 hardware encoding (NVIDIA GPU)
• Quality presets — Generous, Average, and Below Average quality tiers with target bitrate-per-hour calculations
• HQ downscale — download at a higher resolution, then downscale for better quality at lower resolutions

UI & Workflow

• Four-tab layout — Download, Subs, Browse, Settings (Recent now lives under Browse, Index under Settings)
• Settings categories — General / Performance / Appearance / Tools / Index sub-tabs
• Simple / Verbose log modes — toggle between a readable sync view and full yt-dlp output
• Pause / resume — pause active downloads or transcriptions mid-session and resume without losing progress
• GPU task queue — transcription and compression jobs run in their own reorderable queue, separate from downloads
• System tray — sits in the tray with separate indicators for downloads and GPU tasks; auto-sync controllable from the tray menu
• Desktop notifications — Windows notifications on sync completion, errors, etc.
• Internet monitoring — automatically pauses on connection loss, resumes when connectivity is restored
• Drive monitoring — automatically pauses on drive failure, resumes when drive is restored
• Auto-update — checks for new releases on GitHub at startup

Tech Stack

Built on yt-dlp + ffmpeg for downloading, OpenAI Whisper (faster-whisper) for GPU transcription, SQLite FTS5 for the transcript index, Chart.js for graphing, and a pywebview shell rendering an HTML/CSS/JS frontend.

The previous tkinter-based build is archived as the Tkinter legacy release

EXE in releases has everything bundled. If you run from source, you'll need Python 3.13 with pywebview, pystray, and Pillow installed (pip install pywebview pystray pillow). Whisper stays out-of-tree and is invoked via a Python 3.11 venv at runtime.

Line by line, it's roughly 90% Claude-Code, 5% Copilot, 5% me :)