Building Cinematic Videos with AI: Our Video Pipeline Explained

AgenticVexa's video generation isn't just a simple text-to-video model. It's a full cinematic pipeline that produces multi-scene videos with consistent characters, AI narration, dynamic music, and professional transitions.

The Pipeline

When you submit a video prompt, it goes through 7 stages:

1. Story Bible — An LLM creates a comprehensive story document: characters, settings, visual style, and narrative arc
2. Scene Breakdown — The story is divided into individual scenes with specific prompts, camera angles, and timing
3. Image Generation — Each scene is rendered as a high-quality image using FLUX, with character and style consistency maintained across scenes
4. Voice Narration — Kokoro generates natural narration for each scene with appropriate pacing and emotion
5. Music Composition — AI generates a background score that matches the mood and tempo of the video
6. Ken Burns Effects — Subtle pan and zoom animations are applied to each scene image for cinematic motion
7. Final Composition — Everything is assembled with professional transitions, audio mixing, and 4K output

Credits

Video generation costs 50 base credits + 8 credits per scene. A typical 5-scene video costs 90 credits and produces a 60-90 second cinematic piece.

API Usage

job = client.video.generate(
    prompt="A documentary about the history of space exploration",
    scenes=5,
    style="cinematic",
    narration=True,
    music=True
)

# Video generation is async — poll for status
while job.status == "processing":
    time.sleep(10)
    job.refresh()

print(job.video_url)  # Your finished video