Skip to main content

AI Pipeline

The ACT3 AI pipeline connects your production intent to the world's leading AI video generation models.

Supported Models​

Google Veo 3​

Google's most advanced video generation model. Veo 3 produces high-quality, photorealistic video with excellent motion consistency and cinematic lighting. Best for dramatic, narrative content.

Runway Gen-3 Alpha​

Fast, reliable video generation with strong adherence to prompts. Excellent for stylized content, product videos, and quick iteration cycles.

ComfyUI + FLUX​

Local rendering pipeline via ComfyUI with FLUX image and video models. Ideal when you need full creative control, custom LoRAs, or want to keep content on-premises.

Hunyuan Video​

Tencent's open-source video model with strong performance on diverse content types.

Wan 2.1​

High-quality open-source model with excellent temporal consistency for long clips.

How the Pipeline Works​

Shot Definition
    ↓
Prompt Optimization (ACT3 AI Cinematography Layer)
    ↓
Model Selection
    ↓
Generation Request
    ↓
Quality Review
    ↓
Timeline Assembly

Prompt Optimization​

ACT3 AI doesn't just pass your shot description directly to the model. It runs your shot through the Cinematography Layer — a translation system that:

  • Converts your technical parameters into model-specific prompt syntax
  • Applies per-model style modifiers for cinematic quality
  • Injects character consistency tokens when actors are defined
  • Adds temporal consistency hints for multi-shot sequences

Multi-Take Generation​

For important shots, generate multiple takes simultaneously and compare them side-by-side. ACT3 AI's take manager keeps all generations organized and linked to their source shot.