Midjourney Video Generator DOMINATES at These 3 Things

Key Concepts

• AI Video Generation
• Midjourney Video Feature (Animation)
• Auto vs. Manual Animation Modes
• Low Motion vs. High Motion Settings
• Video Extension Feature
• User Interface (UI) and Organization
• Prompt Adherence
• Stylistic Range and Aesthetics
• Moderation
• Aspect Ratios
• Performance Benchmarks (Motion, Emotions, Complex Actions)
• Value (Price & Unlimited Relaxed Generations)

Midjourney Video Feature Overview

The video discusses the newly released video generation feature within Midjourney, emphasizing its strengths, weaknesses, and unique characteristics compared to other AI video generators. The speaker expresses surprise at the quality and fun factor of the tool, having used it to create a full music video and run various stress tests.

Access and Basic Functionality

• The video feature is included in existing Midjourney plans, offering generous usage, even compared to models like V3.
• Users can access the feature by opening existing generated images or uploading their own.
• Four animation options appear at the bottom right of the screen.
• Two primary modes:
  • Auto: Animates based on the existing image and original prompt.
  • Manual: Allows users to input a new prompt to control animation.
• Two motion settings within each mode:
  • Low Motion: Provides dynamic movement while maintaining consistency.
  • High Motion: Adds more energy but can introduce warping and distortion. The speaker usually prefers low motion unless creating a chaotic scene.

Speed and Export

• Midjourney is cited as the fastest AI video generator currently available.
• Previews are generated in 480p but can be exported in full 1080p.
• Users are instructed to use the "download video for social media" button for full quality exports, not the "download raw video" button.

Exploration and Initial Impressions

• The speaker tested the feature by animating old Midjourney generations without new prompts.
• Midjourney can animate images in unique and interesting ways compared to other generators.
• It holds together well even with abstract or surreal concepts.

The Extend Feature

• The extend feature allows users to add 4 seconds to the clip and extend videos up to 21 seconds.
• The extend option is available in both low or high motion.
• Transitions between extended segments are smoother than in other video generators.
• Complex actions can be broken down across multiple extensions, demonstrated by a dragon blowing fire into a cup and a woman taking a sip. Splitting the action into separate extensions with different prompts helped achieve the desired result.

User Interface (UI)

• Midjourney's UI is considered the best among current video generators, praised for being clean, fast, and organized.
• Users can "like" generations and filter results, create folders, and use smart folders for organization.
• Organization features that benefit image generation are carried over to video.

Spotter Studio (Sponsor)

• Spotter Studio is a YouTube ideation suite designed to help creators brainstorm, outline, and optimize their content.
• Key Features:
  • Idea Generation: Provides new video angles based on previous successful videos.
  • Title Workshop: Suggests alternative titles.
  • Thumbnail Creation: Generates thumbnail options based on the video title and topic.
  • Story Beats: Helps outline the video structure with a focus on narrative arc (goal, hook, challenges, climax, resolution).
  • Packaging Preview: Shows how the title and thumbnail will appear on YouTube across different devices and includes a squint test.
  • Outliers Tab: Identifies overperforming videos on other channels to spot trends and generate high-potential ideas.

Abstract World Building

• The speaker experimented with starting with abstract images and using the extend feature repeatedly without new prompts to let Midjourney develop the world organically.

Nick Staint Pierre's Example

• The speaker highlighted an example by Nick Staint Pierre showcasing advanced control over each extension segment using specific prompts. The example features a character following complex motions with a butterfly landing on his nose, demonstrating the platform's potential.

Pricing and Unlimited Generations

• Upgrading to the $60/month plan grants unlimited generations in "relax mode" (slower generation times).
• This plan is considered the cheapest unlimited option compared to other models or local setups.
• Users can queue multiple generations and extensions, making it cost-effective for heavy use.

Uploading Custom Images

• Users can upload their own images to animate.
• The video shows an example of uploading an image created using a custom LoRA model of the speaker.
• Uploaded images can be animated by specifying prompts and adjusting settings like motion and speed.

Tests and Limitations

Hands and Text

• Hands are surprisingly well-rendered in video, despite being a challenge for the image generator.
• Text that is part of the initial image stays consistent during animation, but new text cannot be introduced.

Complex Motions

• Breakdancing is rendered impressively, though not perfectly.
• Gymnastics and cartwheels are still challenging for AI.
• Action sequences (fight scenes, MMA, T-Rex chasing) look coherent but struggle with accurate physics and reactions to impacts.
• Precise movements like knocking over dominoes or threading a needle are not yet mastered.
• Focus control (shifting focus from foreground to background) works surprisingly well.
• Shooting a basketball is still a challenge.

Emotions

• Subtle and basic emotions are handled well.
• Larger emotions in groups (friends laughing, couples arguing) are rendered effectively.
• The model struggles with changing emotions in a single animation, especially when the starting expression conflicts with the target emotion.

Walk and Run Cycles

• Walk cycles are generally good with occasional stutter steps.
• The model can maintain believable motion across extensions, including transitions from walking to running.

Astra by Topaz Labs

• Astra by Topaz Labs, specifically the creative upscaler, is recommended for enhancing AI-generated footage.
• It works well with Midjourney videos, adding detail and cleaning up artifacts, especially when upscaling older Midjourney generations.

Moderation

• Moderation is stricter than in the image model, possibly due to ongoing lawsuits.
• The filter can be overly aggressive, flagging seemingly harmless prompts and images.
• Exposed shoulders and violence are flagged, but the criteria appear inconsistent.

Aspect Ratios

• The video feature supports various aspect ratios, including cinematic widescreen (2.39:1) and extreme aspect ratios (e.g., 10:1).

Stylistic Range

• Midjourney excels in stylistic range, handling illustrative styles, anime, 3D animation, pixel art, hand-drawn sketches, and cartoons effectively.
• The speaker showcases a compilation demonstrating the range from polished and cinematic to stylized and experimental.

Overall Assessment

• Midjourney video is generally good across most areas, on par with other major models.
• It is not the best at complex movements or advanced camera movements.
• It stands out in speed, aesthetics, stylistic range (especially illustrative/abstract), organization, and price (due to unlimited relax mode generations).
• The speaker intends to incorporate it into their workflow, especially since they already subscribe to Midjourney.

Futuredia

• Futuredia offers over 20 comprehensive courses on incorporating AI into various aspects of life and career.
• They offer a 7-day free trial.

Synthesis/Conclusion

Midjourney's new video generation feature is a powerful tool that combines speed, aesthetic quality, and organizational capabilities at a competitive price point. While it may not be the absolute best in every category (particularly complex motion), its strengths in stylistic range, user interface, and unlimited generation option make it a compelling addition to the AI video generation landscape. The "extend" feature is particularly noteworthy for creating seamless transitions and evolving scenes. The discussion also highlighted the importance of tools like Spotter Studio for content creators and Astra for enhancing visual quality, further emphasizing the ecosystem surrounding AI-powered video creation.