AI Video Generation
ACT3 AI transforms your scripts, scenes, and shot prompts into video clips using a suite of AI rendering engines. You choose the engine that fits your style and quality requirements, then direct the output using natural language cinematography instructions.
How It Works
The generative pipeline connects your creative decisions to AI rendering in a continuous loop:
- Script → Structure — Your script is broken into acts, scenes, beats, and shots
- Shot Prompts — Each shot has a text description with camera angle, movement, mood, and subject
- Engine Selection — Choose from Google Veo 3.1, Grok Video, or Wan 2.1
- Generation — ACT3 AI sends the prompt to the engine and returns a video clip
- Review — Preview the clip and accept it, refine the prompt, or regenerate
- Assembly — Approved clips go to the timeline for final editing
Rendering Engines
| Engine | Best For | Output Quality |
|---|---|---|
| Google Veo 3.1 | Photorealistic, cinematic output | Up to 1080p at 30fps |
| Grok Video | Creative and stylistically distinctive video | Multiple resolutions |
| Wan 2.1 | Fast collaborative previews | Low-res preview |
See the Integrations section for detailed documentation on each engine.
Writing Shot Prompts
Shot prompts are the creative instructions that tell the AI what to generate. Strong prompts include:
- Shot type — wide shot, close-up, over-the-shoulder, aerial
- Camera movement — dolly in, tracking shot, handheld, Steadicam, static
- Subject — who is in the shot and what they are doing
- Setting — location, time of day, weather
- Mood and lighting — noir, golden hour, dramatic shadows, cool blue
Example prompts:
- "Low-angle medium shot of detective approaching the door, handheld camera, dramatic shadow lighting, tense atmosphere."
- "Wide establishing aerial shot of futuristic city at night, drone movement, neon reflections on wet streets."
- "Extreme close-up of astronaut's eyes, shallow depth of field, helmet reflection shows approaching planet, tense."
Keep prompts focused and under 100 words. Use standard film terminology for best results.
Render Styles
Render Styles define the overall visual aesthetic — color grading, film grain, contrast, and mood — applied at render time to give your entire project a unified cinematic look.
AI Directing
The AI Directing feature translates high-level creative decisions into shot-level settings:
- Narrative Pacing Control — Set tempo for a scene (slow burn, fast cuts, balanced)
- Tone and Mood — Apply consistent lighting, color, and emotional style across scenes
- Cinematic Presets — Built-in looks like Film Noir, Wes Anderson, Sci-Fi Epic, Documentary
- Continuity Matching — Maintain consistent lens choice, exposure, and framing style across related shots
- Preview Variations — Generate multiple visual takes before committing credits
To use AI Directing: open a scene or shot, click AI → Directing, set your parameters, and preview before approving.
The Generative Workflow Pipeline
A complete generative project moves through these stages:
- Ideation — Start a new project with a concept, format, and aspect ratio
- Script Development — Write or import, then expand with AI Wizard
- Characters and Casting — Define characters and assign digital actors
- Visual Planning — Top-Down View and set design
- Generative Rendering — Generate draft previews for each shot
- Assembly and Editing — Combine clips, B-roll, and audio in the timeline
- Export — Deliver final video in the format and resolution you need
Preview vs Final Rendering
Use draft previews for all creative iteration. Only commit to final quality renders for approved content:
| Quality | Use Case | Credit Rate |
|---|---|---|
| Draft | Creative review, timing checks | Lowest |
| 720p | Client preview, rough cut sharing | Low |
| 1080p | Final delivery for most platforms | Standard |
| 4K | Premium output, large screens | Highest |
Draft renders using Wan 2.1 are the most cost-effective way to test ideas before committing to a full-quality render.
Build Video
Once all your shots are generated and approved, use the Build Video feature to assemble the final production:
- Open Build Video mode from the editor
- Drag scenes, shots, and B-roll into the timeline
- Add transitions, audio tracks, and music
- Apply color grading for visual consistency
- Preview the full sequence
- Send to the Render Queue for final export
Build Video supports MP4, MOV/ProRes, and WebM. Aspect ratios include 16:9, 9:16, and 1:1 for different platforms.
Credit Usage Examples
| Output | Engine | Approximate Credits |
|---|---|---|
| 30-second clip at 1080p | Google Veo 3.1 | ~180 credits |
| 2-minute short film preview | Wan 2.1 draft | ~80 credits |
| 2-minute short film final | Runway 1080p | ~600 credits |
| 4K 10-second shot | Hunyuan | ~10 credits |
Always check the credit estimate shown before confirming a render.
Best Practices
- Use draft previews for creative iteration — save high-quality renders for final output
- Break complex shots into shorter clips rather than trying to generate long sequences
- Use Wan 2.1 for team collaboration and real-time idea testing
- Apply consistent cinematographic style across all shots for a cohesive look
- Batch similar jobs in the Render Queue to avoid bottlenecks
AI Role System
ACT3 AI breaks the filmmaking process into specialized AI roles that work together:
AI Writer — Expands high-level ideas into full scripts. Give it a logline or a paragraph and it generates a complete beat sheet, scene outline, and dialogue draft. You can also use it to rewrite individual scenes or fill in missing dialogue.
AI Director — Orchestrates your overall creative vision. It manages scene pacing, story flow, and the sequencing of shots to ensure the narrative reads correctly. It works alongside the AI Writer and AI Cinematographer to propose a cohesive interpretation of your script.
AI Cinematographer — Translates your story beats into camera decisions: shot type, lens focal length, camera movement, and framing. It draws from 22 standard shot types and key-framed camera curves to propose cinematography for each beat automatically.
AI Set Designer — Creates virtual environments for your scenes. Based on your script context and story location, it suggests or generates appropriate sets — from procedural city blocks to interior rooms. You can accept suggestions, modify them, or build from scratch.
Each AI role makes suggestions that you can accept, override, or regenerate. The human creator remains in control at every step.
The Mega Prompt
The Mega Prompt is the full assembled instruction packet ACT3 AI sends to the generative video engine for each shot. It bundles your narrative description, style preset, camera settings, lighting setup, audio cues, and motion data into one multi-modal prompt.
You can view and edit the Mega Prompt directly in the AI Prompt Editing panel for expert-level control. Click Copy Prompt to copy the full assembled prompt to your clipboard for debugging, reference, or use in other tools.
Shot Rendering Modes
Each shot can use a different rendering pipeline:
| Mode | Description | Best For |
|---|---|---|
| Generative AI Only | Pure AI video generation | Most shots, all styles |
| 3D Characters on 2D Background | Blender character over flat plate | Character accuracy on simple backgrounds |
| Full 3D Environment | Complete Blender render | High-fidelity 3D scenes |
| Hybrid | 3D characters composited over AI background | Action sequences, existing 3D assets |
Set the rendering mode per shot in the master dropdown. Hybrid mode is useful when you need precise character placement (from Blender) composited over an AI-generated background.
Slow Motion and Frame Rate
To generate slow motion, set the frame rate override on a shot to high-frame-rate and include slow motion language in your shot prompt (for example: "slow-motion close-up"). The video editor can also apply time remapping to any generated clip to create slow motion in post.
Repeatable Results with Seed Values
Use a fixed seed value in the advanced shot settings to make generation results more predictable. A seed locks the randomness in the model, producing very similar results when you regenerate with the same prompt and seed. This is useful for making minor prompt adjustments while preserving the core look of a shot.
Regenerating and Version History
Click the regenerate icon on any shot to generate a new version. You can adjust the shot prompt, camera settings, or model before regenerating. Previous versions are kept in the shot's version history so you can compare and revert at any time.
Multi-Character Shots
To generate a scene with two or more characters interacting, assign all relevant digital actors to the shot in the Cast panel. In your shot prompt, describe the interaction between them by name (for example: "Sarah and David argue across the kitchen table, Sarah pointing at him"). ACT3 AI passes all actor references to the model to maintain both appearances.