AI Video Generation

Quick Reference

Model Selection

# Auto-select based on complexity
from app.sora_helper import generate_video_smart

result = await generate_video_smart(prompt, topic, duration=8)
# Returns: {success, video_path, model_used}

Complexity Keywords:

• HIGH (Sora Pro): transformation, morphing, cinematic, epic
• MEDIUM (Sora): tracking, dolly, movement, transition
• LOW (Veo): static, simple, product

Prompt Formula

[CINEMATOGRAPHY] + [SUBJECT] + [ACTION] + [CONTEXT] + [STYLE]

Example:

Slow dolly shot, IoT sensor on greenhouse shelf, LED blinking green,
morning sunlight through glass, professional documentary style

Generation Functions

# Veo (Google)
from app.veo_helper import generate_video_veo3
result = await generate_video_veo3(prompt, aspect_ratio="9:16", duration_seconds=8)

# Sora (OpenAI)
from app.sora_helper import generate_video_sora
result = await generate_video_sora(prompt, model="sora-2", size="720x1280")

# Kling (fal.ai)
from app.fal_helper import FalVideoGenerator
result = await FalVideoGenerator.generate_video(prompt, model="kling_pro", duration=10)

Instagram Specs

Environment Variables

GEMINI_API_KEY=...   # Veo
OPENAI_API_KEY=...   # Sora
FAL_API_KEY=...      # Kling

Return Format

# Success
{"success": True, "video_path": "/path/to.mp4", "model_used": "veo-3.1"}

# Error with fallback
{"success": False, "error": "...", "fallback": "veo3"}

Best Practices

• Simple, comma-separated sentences
• Include camera movement + lighting
• Avoid complex physics (bouncing, running)
• Use: “cinematic”, “professional”, “documentary”
• Duration: 5-8s ideal for Reels

Deep Links

• app/sora_helper.py – Sora + smart selection
• app/veo_helper.py – Veo generation
• app/fal_helper.py – Kling/Wan/Minimax
• app/video_models.py – Model configs
• context/reels-prompts.md – Prompt examples