Phase 4 — Videos

Phase 4 generates individual high-definition video clips for every scene, complete with synchronized narration or dialogue, immersive sound effects, and camera motion.

1. Top Controls and Video Models

The header controls dictate video models, resolution, voice styling, and camera transitions:

2. Audio Tracks and Voice Synthesis

Audio options provide flexibility for standalone publishing or post-production workflows:

• Full Audio (Default): Generates natural vocal delivery alongside contextual sound effects and foley.
• SFX Only: Characters stay silent while ambient sound and foley remain intact, ideal for importing studio voice recordings later.
• Mute: Generates silent video clips for custom music tracks and sound design in external video editors.
• Narrator Profiles: Deep Documentary (warm masculine baritone, ~130 wpm), Serene Documentary (clear feminine contralto, ~135 wpm), Storyteller (mature reflective masculine, 48yo), or Dynamic Youth (clear modern timbre, ~145 wpm).
• Custom Voices: Define custom voice profiles (15 to 400 characters) specifying age, gender, accent, tone, and pacing.

3. Scene Video Cards and Prompt Assembly

Each scene card displays the rendered clip alongside its input parameters:

• Storyboard Preview Anchor: Displays the initial keyframe used as the mandatory first frame of the video.
• Synchronized Dialogue: Displays the spoken text and the assigned voice profile. In NARRATIVE scenes, the engine requests natural lip-sync matching the audio.
• Prompt Architecture: Merges concise style rules, framing directives, camera movement, character identity anchors, audio cues, and speech pacing constraints.
• Negative Exclusions: Automatically excludes watermarks, floating UI elements, split-screen artifacts, and distorted anatomy.

4. Continuity Modes Across Scenes

Continuity options govern how adjacent scenes connect visually:

• Independent: Each scene starts strictly from its own storyboard keyframe without connecting to neighboring scenes.
• Continuous Video (Anchor Next) — Default: Uses the next scene's storyboard as the last frame of the current scene, causing the camera action to naturally land where the next scene begins.
• Inherit Previous Last Frame: Extracts the actual rendered last frame from the previous video clip and uses it as the starting frame of the subsequent clip.

5. Automated Technical Constraints

The tool automatically formats prompts to match model-specific limitations:

• Duration Matching: Omni 1.1 Flash accepts 4, 6, 8, or 10-second clips. Veo 3.1 models operate strictly on 8-second clips. The engine chooses the shortest valid duration that accommodates the spoken word count.
• Reference Rules: Google Flow accepts up to 5 reference images per video clip and treats initial frames and direct references as mutually exclusive. The engine compiles these configurations automatically.