Transform spoken content into high-converting video ads — in seconds.
An AI-driven pipeline that ingests audio/video transcripts, analyses them with NLP, and automatically generates ad scripts, storyboards, and MP4 video previews.
Transcript-to-Ad Generator is a production-ready, end-to-end pipeline for converting raw transcripts into polished advertising content. It combines spaCy NLP for intelligent content extraction, Ollama (local LLM) for creative ad-copy generation via structured prompting, MoviePy for on-the-fly video rendering, and Redis Queue for asynchronous background processing — all wrapped in a premium Streamlit UI.
| Feature | Description |
|---|---|
| 🧠 Deep NLP Analysis | spaCy + TextBlob extract keywords, named entities, sentiment, and highlights |
| ✍️ Dual-Mode Ad Generation | Ollama LLM mode for creative copy · Simple mode for fast deterministic output |
| 🎞️ Visual Storyboarding | Auto-maps ad copy to visual frames with scene descriptions |
| 🎬 Video Preview Rendering | Generates MP4 previews on-the-fly via MoviePy |
| ⚡ Async Background Jobs | Redis Queue (RQ) workers handle heavy rendering without blocking the UI |
| 🖥️ Premium Web UI | Dark-mode Streamlit interface with custom CSS and real-time feedback |
| 💾 Smart Caching | Caches expensive LLM and rendering calls for speed |
| 🐳 Production Ready | Fully Dockerized with multi-container compose, pre-commit hooks, and structured logging |
This project uses Ollama to run large language models locally — no API key, no cost, no data leaving your machine. The core/llm.py module handles structured prompt construction and response parsing.
# 1. Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh
# 2. Pull a model (Mistral recommended)
ollama pull mistral
# 3. Start the Ollama server
ollama serveOnce running, the pipeline connects to http://localhost:11434. Switch models by setting OLLAMA_MODEL in your .env file.
| Layer | Technology |
|---|---|
| NLP | spaCy · TextBlob |
| LLM / Prompt Generation | Ollama (local) — Mistral / Llama 3 |
| Video Rendering | MoviePy · Pillow · FFmpeg |
| Async Workers | Redis · RQ (Redis Queue) |
| Web UI | Streamlit (custom CSS, dark mode) |
| Backend | Python 3.9+ |
| Containerisation | Docker · Docker Compose |
| Quality & CI | pytest · flake8 · mypy · pre-commit · GitHub Actions |
Transcript-to-Ad-Generator/
├── .github/workflows/ # CI/CD — GitHub Actions pipelines
├── core/
│ ├── analysis.py # NLP & text analysis (spaCy, TextBlob)
│ ├── ad_generator.py # Ad copy generation logic
│ ├── llm.py # Ollama LLM integration & prompt generation
│ ├── nlg.py # Natural language generation helpers
│ ├── video.py # Video rendering (MoviePy)
│ ├── pipeline.py # Main orchestration layer
│ ├── gap_scanner.py # Opportunity & gap detection
│ └── queue.py # Redis Queue interface
├── scripts/ # Worker & Redis utility scripts
├── tests/ # Unit & integration tests (pytest)
├── app.py # Streamlit web application
├── run_local.py # CLI runner for the pipeline
├── utils_logging.py # Structured logging setup
├── Dockerfile # App container
├── Dockerfile.worker # RQ worker container
├── docker-compose.yml # Full stack compose (app + worker + Redis)
├── Makefile # Dev shortcuts
├── pyproject.toml # Project metadata & tool config
├── mypy.ini # Type checking config
├── .flake8 # Linting config
├── .pre-commit-config.yaml # Pre-commit hooks
├── .env.example # Environment variable template
└── requirements.txt # Python dependencies
- Python 3.9+
- FFmpeg — required for local video rendering
- Ollama — for LLM-powered ad generation
- Docker — optional, recommended for full stack
git clone https://github.com/ImdataScientistSachin/Transcript-to-Ad-Generator
cd Transcript-to-Ad-Generator
python -m venv .venv
# Windows
.\.venv\Scripts\Activate
# Linux / macOS
source .venv/bin/activate
pip install -r requirements.txtcp .env.example .env
# Edit .env — set OLLAMA_MODEL, Redis connection, etc.streamlit run app.pyOpen http://localhost:8501 in your browser.
Spin up the app, Redis, and the background worker in a single command:
docker-compose up --build| Service | Port | Description |
|---|---|---|
| Streamlit App | 8501 |
Main web interface |
| Redis | 6379 |
Job queue backend |
| RQ Worker | — | Background video renderer |
# Run the full test suite
pytest tests/
# Type checking
mypy core/
# Linting
flake8 core/
# Run all pre-commit hooks manually
pre-commit run --all-filesCI/CD via GitHub Actions runs linting, type checks, and tests automatically on every push.
Run the full pipeline directly from the command line without the UI:
python run_local.pyProcesses a sample transcript and prints the NLP analysis and generated ad copy to the console — useful for scripting and batch workflows.
| Item | Status | Notes |
|---|---|---|
| GitHub Actions CI/CD | ✅ Done | .github/workflows/ confirmed present |
| Ollama LLM Integration | ✅ Done | core/llm.py — local prompt generation |
| Docker + Compose | ✅ Done | App + worker + Redis fully composed |
| Redis Queue workers | ✅ Done | Dockerfile.worker + core/queue.py |
| Code quality tooling | ✅ Done | flake8, mypy, pre-commit all configured |
| README quality | ✅ Done | This document |
| Repo name (trailing dash) | ❌ Pending | Original name transcript_To_Ad_Generation- — rename in GitHub Settings |
| GitHub About description | ❌ Pending | No sidebar description set — add via ⚙️ repo Settings |
| Topics | Add: ollama github-actions moviepy — verify generative-ai is present |
Contributions are welcome!
- Fork the repository
- Create your feature branch —
git checkout -b feature/your-feature - Commit your changes —
git commit -m 'Add your feature' - Push to the branch —
git push origin feature/your-feature - Open a Pull Request
Distributed under the MIT License — free to use, modify, and distribute.
See LICENSE for details.
Sachin Paunikar
