YouTube Video Summarizer

Paste a YouTube URL → the app downloads the audio, transcribes it with local Whisper, then summarizes it with any text model on OpenRouter.

No YouTube captions required.

Repo: nsdivyasingh/Youtube-Video-Summerizer

---

Local setup

1. Install ffmpeg and put it on your PATH.
2. Install Python deps:

   pip install -r requirements.txt

3. Create a .env file (copy from the example) and add your OpenRouter key:

   cp .env.example .env

   OPENROUTER_API_KEY=your-openrouter-api-key-here
   OPENROUTER_MODEL=meta-llama/llama-3.3-70b-instruct

4. Run:

   streamlit run app.py

Get a key at openrouter.ai/keys.
.env is gitignored — never commit real API keys.

---

Deploy on Streamlit Community Cloud

The in-app Deploy button often fails for private repos or OneDrive paths. Deploy from the website instead:

1. Make sure main is on GitHub (already pushed to
   https://github.com/nsdivyasingh/Youtube-Video-Summerizer).
2. Open share.streamlit.io and sign in with GitHub.
3. Click New app and choose:
   • Repository: nsdivyasingh/Youtube-Video-Summerizer
   • Branch: main
   • Main file path: app.py
4. Under Advanced settings → Secrets (or Manage app → Settings → Secrets), paste:

   OPENROUTER_API_KEY = "your-openrouter-api-key-here"

5. Click Deploy, then reboot the app after saving secrets. If the repo does not appear in the list:

• Grant Streamlit access to the repo (GitHub → Settings → Applications → Streamlit), or
• Make the repo Public (GitHub → Settings → Danger zone → Change visibility).

packages.txt installs ffmpeg on Cloud. First run downloads the Whisper model and can take a few minutes.

---

How it works

1. yt-dlp downloads audio (up to 15 minutes)
2. faster-whisper (tiny) transcribes locally
3. OpenRouter returns the summary

---

Notes

• OpenRouter audio/video models need account credits; this app only uses OpenRouter for text summary.
• Free Cloud apps have limited RAM. If the app is killed during transcription, try a shorter video or run locally with a larger Whisper model.