I Tested The Most Insane AI Video Prompts to See What's Possible

AI Video Generation: Pushing the Limits

Key Concepts: AI video generation, Runway Gen 4, Kling V2, Higsfield, object permanence, character persistence, complex camera movements, prompt engineering, limitations of AI video generation, AI video editing, AI video tools.

Object Permanence and Character Persistence

• Object Permanence: AI models, particularly Gen 4, have significantly improved in maintaining object consistency even when occluded. Example: Reflection of a burning building in an eye remains consistent even when the eye closes.
• Character Persistence: Facial features and character appearance remain consistent throughout the shot, a significant improvement over previous models.
• Example: A clown waiting for a bus remains the same clown even after a car passes in front of him.

Emotional Expression and Multiple Characters

• AI models are now better at generating videos with characters displaying a range of emotions.
• Models can handle multiple characters interacting in a scene.
• Example: Multiple characters laughing together.

Complex Scenes and Styles

• AI can now generate complex scenes, including reflections, focus shifts, and diverse artistic styles.
• Example: Reflection of someone playing the piano.
• Example: Paper airplane flying out of a window with a coherent new scene outside.
• Example: Focus shift between foreground and background characters with movement.
• Models can combine styles, such as realistic restaurant with a 2D cartoon waitress.

Spotter Studio: AI-Powered YouTube Tool (Sponsored)

• Spotter Studio is a tool for YouTube strategy, ideation, and design.
• Brainstorming: Generates title and thumbnail ideas based on a starting point.
• Outliers Tab: Identifies high-performing videos in a specific niche or on a particular channel to understand audience interests.
• Audience Also Watched Tab: Finds videos that resonate with a specific audience.

Complex Camera Movements

• Whip Pan: Higsfield's new update with preset shot types can achieve whip pan camera movements, which were previously difficult.
• Action Shot: Higsfield also excels at fast-moving action scenes using preset shot types.
• Example: Woman running from a T-Rex.

Selective Coloring and Specific Actions

• Gen 4 can perform selective coloring, such as a black and white scene with red lipstick.
• Gen 4 can generate specific actions, such as blowing out candles.
• Example: Woman adding lipstick to her top lip.

Limitations and Challenges

• Domino Effect: AI struggles to accurately simulate the domino effect.
• Breakdancing: While not perfect, AI can generate passable breakdancing scenes with editing.
• Juggling: AI struggles with complex juggling scenarios, especially with additional elements like a clown on a unicycle.
• Throwing: Basic throwing actions, like tossing keys, are difficult to achieve realistically.
• Basketball Shot: Consistently generating a realistic basketball shot into a hoop is nearly impossible.
• Reverse Environments: AI struggles with scenes where a character walks normally while the environment moves in reverse.
• Fine Motor Skills: Precise movements, such as threading a needle, are difficult to generate.
• Multiple Actions: Scenes involving multiple actions, especially with multiple characters, are challenging.
• Example: A dragon blowing flame into a cup and then taking a sip.
• Fight Scenes: Often result in slow motions or blurry movements.
• Upside Down: Scenes involving people going upside down, such as gymnastics or cartwheels, are very difficult.

Notable AI Video Examples

• Uncanny Harry: Bullet time shot in Higsfield combined with Luma for a full scene.
• Runway CEO: Underwater shot with realistic physics and splashes.
• Runway: Full animated short film generated entirely in Gen 4.
• Sway Molina: Surreal bird people animations.
• Alex Petrascu: Glass and wood assembling animations.
• Jav Lopez: Realistic skateboarding shot.

Conclusion

AI video generation has made significant strides, achieving realistic scenes, character consistency, and complex camera movements. However, limitations remain in simulating precise actions, complex physics, and multi-character interactions. Despite these challenges, AI video generation is rapidly evolving, enabling the creation of short films and realistic videos. Futureidia offers courses on incorporating AI into various aspects of life and career.