Complete Guide: Creating a Multi-Camera Talking Head Video in Google Flow

Complete Guide: Creating a Multi-Camera Talking Head Video in Google Flow
This workflow is designed specifically for Google Flow. The objective is to transform a single-camera talking head video into a cinematic multi-camera edit while preserving the speaker’s identity, voice, lip sync, lighting, and environment exactly as in the original.

Step 1: Record the Source Video
For the best results in Google Flow, your source video should have:

  • Resolution: 1080p or 4K
  • Stable camera (tripod recommended)
  • Good lighting
  • Clean background
  • Speaker facing the camera
  • Minimal body movement
  • Clear, high-quality audio
  • Continuous talking without jump cuts
    Avoid:
  • Camera shake
  • Fast hand movements
  • Motion blur
  • Frequent head turns
  • Poor lighting

Step 2: Upload the Video to Google Flow
Upload your original talking head video without trimming or editing it.
The original video will serve as the master reference for identity, lip sync, and environment.

Step 3: Use Gemini 2.5 Pro (Highly Recommended)
Before generating the video, ask Gemini 2.5 Pro inside Google Flow to create an optimized prompt for your specific video.
Example:
Analyze this talking-head video and generate a Google Flow prompt that converts it into a professional multi-camera interview while preserving identity, lip sync, room layout, lighting, and audio exactly.
Gemini usually produces prompts that work better than manually written prompts because they are optimized for Google’s video generation models.

Step 4: Use This Prompt
Maintain the original room, environment, lighting, and the person’s identity exactly as they are. Do not alter the face, facial features, hairstyle, clothing, body proportions, skin texture, eye color, room layout, background objects, or overall appearance. The person and the environment must remain perfectly consistent throughout the entire video.

Do not modify the dialogue, voice, speech timing, pacing, lip sync, facial performance, emotion, or audio in any way. The spoken content must remain identical to the source. Do not add pauses, remove words, or change the expression. The only changes allowed are the virtual camera framing and camera angle.

Use only fast, clean hard cuts between camera angles. Never morph, blend, interpolate, zoom, animate, or create AI transition effects. Every camera change must feel like an instant cut between multiple professional cameras recording the same scene simultaneously. Maintain perfect continuity across every shot.

Shot sequence:

  • At 00:00 to 00:01.2:
    Original medium shot. Keep the framing identical to the source.
  • At 00:01.2 to 00:02.4:
    Hard cut to a cinematic left-side 45° angle. Keep the subject naturally speaking.
  • At 00:02.4 to 00:03.6:
    Hard cut to a slightly high-angle medium close-up.
  • At 00:03.6 to 00:04.8:
    Hard cut to a cinematic right-side 45° angle.
  • At 00:04.8 to 00:06.0:
    Hard cut to an extreme close-up of the eyes and upper face.
  • At 00:06.0 to 00:07.2:
    Hard cut to a slightly low-angle close-up.
  • At 00:07.2 to 00:08.6:
    Hard cut to an over-the-shoulder framing from behind the subject’s left side.
  • At 00:08.6 to 00:10.0:
    Hard cut back to the original camera position and framing.

Critical requirements:

  • Preserve the person’s identity with pixel-level accuracy.
  • Preserve clothing, hairstyle, facial proportions, and posture.
  • Preserve the room exactly.
  • Preserve lighting, shadows, reflections, and colors.
  • No changes to audio.
  • No changes to speech timing.
  • No changes to lip sync.
  • No new gestures.
  • No body modifications.
  • No environment changes.
  • Only the camera position changes.
  • Every shot must appear as if it was captured simultaneously using multiple professional cinema cameras.