YouTube Prep Tools

A self-hosted web app for prepping YouTube videos: cuts silence and exports an EDL for your editor (DaVinci, Kdenlive, etc.), and transcribes audio with Whisper for subtitles or chapter notes.

Originally a Tkinter desktop tool by RavinMaddHatter; this fork is a containerized Streamlit web app you can run on a home server (TrueNAS, Proxmox, any Docker host) and access from a browser.

What it does

• Silence cutter — Analyzes audio levels per channel, removes silent gaps, and exports a .edl you can drop straight into your NLE timeline.
• Whisper transcriber — Transcribes audio with faster-whisper (CTranslate2 backend, runs on CPU). Supports tiny, base, small, medium, large-v3.
• Audio preprocessing — Optional ffmpeg pipeline (highpass + loudnorm, optional aggressive denoise) for rough source recordings.
• Combined folder transcripts — Stitch all transcripts in a folder into one file with optional chapter headers per source clip.
• Server-side file access — Point it at a mounted /data directory; no need to upload multi-GB files through the browser.
• Browser uploads — Up to 4 GB per file for smaller clips.

Quick start (Docker)

Pull and run from GitHub Container Registry:

docker run -d --name yptools --restart unless-stopped \
  -p 8501:8501 \
  -v /path/to/your/videos:/data \
  ghcr.io/rivasmario/youtubepreptools:latest

Then open http://localhost:8501. Your videos will appear at /data/... inside the app.

Deploying on a home server

TrueNAS SCALE

1. Apps → Discover → Custom App
2. Image: ghcr.io/rivasmario/youtubepreptools:latest
3. Port mapping: container 8501 → host 8501
4. Storage: host path of your video dataset → container path /data
5. Save and start, then browse to http://<truenas-ip>:8501

Proxmox

Run inside an LXC or VM with Docker installed:

docker run -d --name yptools --restart unless-stopped \
  -p 8501:8501 -v /path/to/videos:/data \
  ghcr.io/rivasmario/youtubepreptools:latest

Mounting a remote SMB/NFS share into the container

On a Linux Docker host (e.g. via Tailscale to your NAS):

sudo mount -t cifs //192.168.0.203/winset /mnt/winset -o username=nasuser
docker run -d --name yptools -p 8501:8501 -v /mnt/winset:/data \
  ghcr.io/rivasmario/youtubepreptools:latest

Using the app

Sidebar settings

• Audio Thresholds (dB) — enable channels and set the silent-cutoff level per channel.
• Timing Rules — lead-in, lead-out, minimum clip length, minimum silent duration.
• Whisper Model — pick model size. large-v3 is best quality but slowest.
• Audio Preprocessing
  • Normalize + clean audio — safe for almost everything; fixes loudness and rumble.
  • Aggressive denoise — only for muffled/noisy recordings; can hurt clean audio.
• Folder Transcription
  • Combine into one file — stitches all clip transcripts into a single file at the folder level.
  • Add chapter headers — inserts ## filename markers between sections.

Workflow

1. Set your audio thresholds and timing rules in the sidebar.
2. Video Cutter tab: enter a server path (e.g. /data/my-project/) or upload files.
3. Click Cut Single Clip (EDL) or Cut Folder (Merge to EDL) → download the .edl.
4. Import the EDL into DaVinci Resolve, Kdenlive, or any NLE that reads CMX3600.
5. Audio Transcriber section: pick a model and transcribe a single file or a whole folder.

Building from source

git clone https://github.com/RivasMario/YoutubePrepTools.git
cd YoutubePrepTools
docker build -t youtubepreptools:latest .
docker run -d -p 8501:8501 -v /path/to/videos:/data youtubepreptools:latest

Local dev (no Docker)

Requires Python 3.11+ and ffmpeg on the host.

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
streamlit run app.py

Project layout

File	Purpose
app.py	Streamlit web UI
cutter.py	Silence-detection and EDL generation
openai_translator.py	Whisper transcription wrapper (faster-whisper)
access_ffprobe.py	ffprobe metadata parser
aws_translator.py	AWS Translate helper (optional)
youtubePrep.py	Legacy Tkinter desktop entrypoint (kept for reference)
Dockerfile	Container image definition
init.sh	Container entrypoint (starts Streamlit)
requirements.txt	Python dependencies
youtubeDescription.json	Saved settings + boilerplate description text

Credits

Original project: RavinMaddHatter/YoutubePrepTools

• Discord
• YouTube
• ko-fi — tip the original developer