How to Cast AI Actress Tilly Norwood (AI Consistent Character Tutorial)

Overview

This video tutorial and review focuses on how to cast and integrate an AI-generated actress, Tilly Norwood, into AI-created cinematic scenes. The example used is placing her in a scene from The Last of Us. The creator explores multiple AI tools and workflows to achieve a realistic and consistent AI character performance.

---

Key Technological Concepts and Tools

AI-generated Characters

• Tilly Norwood is an AI-generated actress, promoted by her creators as a potential future Hollywood star.
• This raises controversy about AI potentially replacing human actors.

Image and Video Generation Tools

• Nano Banana Used to modify reference photos with prompts (e.g., changing outfits, backgrounds). Produces stylized but somewhat artificial, airbrushed results resembling video game art rather than realistic cinematic images.

• ChatGPT (with image upload capability) Attempts similar image edits using prompts but results still appear fake with grainy textures.

• VU (reference-to-image and reference-to-video) The main and preferred tool in this tutorial for generating cinematic images and videos of AI characters.

  • Allows uploading multiple reference images to improve consistency and realism.
  • Supports cropping and background integration.

Prompt Engineering

• Careful crafting of text prompts is crucial to approach a cinematic look.
• Includes specifying clothing, environment, lighting, and mood.

Background and Scene Composition

• Attempts to recreate a The Last of Us style background.
• Integrates Tilly into the scene by removing original actors and generating new backgrounds.

Video Workflow

• Uses still images generated in VU to create consistent video frames.
• Enables dialogue and movement for AI characters.

Frame Capture and Upscaling

• Captures frames from generated videos to use as references for further video generation.
• The creator prefers to avoid upscalers due to quality loss.

Challenges and Limitations

• Producing high-quality, realistic AI video content for storytelling is complex and difficult.
• AI video tools are often marketed as easy, but achieving specific cinematic results requires multiple steps and iterations.

---

Product Features and Workflow Highlights

1. Collect multiple reference photos of the AI character.
2. Use Nano Banana and ChatGPT for initial image modifications (outfit, background).
3. Employ VU’s reference-to-image and reference-to-video features for better cinematic results.
4. Use prompt sidekick tools (custom prompt generators) to refine prompts for style matching.
5. Remove unwanted elements from backgrounds and generate new backgrounds.
6. Capture frames from videos to create consistent reference images for continuous video generation.
7. Adjust and crop images to focus on character faces for better consistency.
8. Combine character images with environment backgrounds and generate video sequences.
9. Add dialogue by creating talking frames using VO or WAN tools.
10. Emphasize that AI-generated cinematic video is still challenging and requires technical skill.

---

Analysis

• The creator expresses skepticism about AI actors replacing humans in Hollywood but embraces AI for independent, amateur filmmaking.
• The tutorial is practical, showing trial and error with different tools and highlighting their strengths and weaknesses.
• The creator’s favorite method is VU’s reference-to-video approach, which provides the best balance of realism and control.
• The video serves as both a guide and a realistic review of current AI video generation capabilities.

---

Main Speakers and Sources

• Video Creator (unnamed): Provides commentary, demonstrations, and personal opinions throughout the tutorial.
• Tilly Norwood: The AI-generated character featured (not a speaker).
• AI Tools Mentioned: Nano Banana, ChatGPT, VU, VO/WAN (voice-over or animation tools).