AI Content Generator

Text-to-Image, Text-to-Video, and Text-to-Speech in a single pipeline — runs on Apple Silicon (MPS), NVIDIA GPU (CUDA), or free Google Colab T4.

---

What It Does

Generate rich media from text prompts — images, videos, and narrated audio — all from one Gradio interface. Designed to run free on Google Colab (T4 GPU) or locally on Mac M1/M2/M3 with Apple MPS acceleration.

---

Capabilities

Mode	Model	Speed	Quality
Text-to-Image	SDXL-Turbo (Stability AI)	~2s (MPS)	High
Text-to-Video	ModelScope text-to-video	~30s (T4)	Medium
Text-to-Speech	Bark (Suno)	~10s	High quality
Text-to-Speech	Edge TTS (Microsoft)	~1s	Fast, free
Combined	Video + Audio narration	—	Full pipeline

---

Architecture

Text Prompt
     │
     ├──▶ ImageGenerator  ──▶ SDXL-Turbo (stabilityai/sdxl-turbo)
     │                         AutoPipelineForText2Image
     │                         Apple MPS / CUDA / CPU
     │
     ├──▶ VideoGenerator  ──▶ ModelScope (damo-vilab/text-to-video-ms-1.7b)
     │                         TextToVideoSDPipeline
     │
     ├──▶ AudioGenerator  ──▶ Bark (suno/bark) — high quality
     │                    ──▶ Edge TTS          — fast, free
     │
     └──▶ Pipeline        ──▶ Combined: video + audio → final output

---

Quick Start

Option 1: Google Colab (Recommended — Free T4 GPU)

1. Open notebooks/AI_Content_Generator_MVP.ipynb
2. Upload to Google Colab
3. Runtime → Change runtime type → T4 GPU
4. Run all cells → Gradio public URL is generated automatically

Option 2: Local (Mac M1/M2/M3)

git clone https://github.com/apuroopy1-prog/ai-video-generator.git
cd ai-video-generator

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

python app.py
# Open http://localhost:7860

---

Project Structure

ai-content-generator/
├── app.py                          # Main Gradio web UI
├── quick_start.py                  # Minimal Python API example
├── requirements.txt
├── notebooks/
│   └── AI_Content_Generator_MVP.ipynb   # Colab notebook
├── src/
│   ├── image_generator.py          # SDXL-Turbo text-to-image
│   ├── video_generator.py          # ModelScope text-to-video
│   ├── audio_generator.py          # Bark + Edge TTS
│   └── pipeline.py                 # Combined video+audio pipeline
├── configs/                        # Model config files
├── huggingface-space/              # HuggingFace Spaces deployment
└── outputs/                        # Generated files (gitignored)

---

Python API

from src.image_generator import ImageGenerator
from src.video_generator import VideoGenerator
from src.audio_generator import AudioGenerator

# Generate image
img_gen = ImageGenerator()
image = img_gen.generate("A futuristic city at sunset, cyberpunk style")

# Generate video
vid_gen = VideoGenerator()
video = vid_gen.generate("A robot walking through a forest")

# Generate speech
audio_gen = AudioGenerator()
audio = audio_gen.generate("Welcome to the future of AI content creation")

---

Hardware Requirements

Hardware	Image	Video	Audio
Apple M1/M2/M3 (MPS)	✅ Fast	✅ Slow	✅
NVIDIA GPU (CUDA)	✅ Fast	✅ Fast	✅
Google Colab T4	✅ Fast	✅ Fast	✅
CPU only	✅ Very slow	⚠️ Very slow	✅

---

Tech Stack

Component	Technology
Text-to-Image	Diffusers, SDXL-Turbo, AutoPipelineForText2Image
Text-to-Video	ModelScope, TextToVideoSDPipeline
Text-to-Speech	Bark (suno), Edge TTS
UI	Gradio
Deep Learning	PyTorch (MPS + CUDA)
Models	HuggingFace Hub

---

Built By

Apuroop Yarabarla — AI/ML Engineer & AI Product Owner