GEN-3: The Ultimate Prompting Guide

Theoretically Media
1 Jul 202411:54

TLDRThe video explores the advancements in Runway ML's Gen 3, a significant upgrade from its predecessor, marking a new era in AI video generation. The host provides an in-depth guide on effective prompting techniques for Gen 3, showcasing its ability to understand descriptive prompts better than keyword spamming. Examples and community ideas are shared, demonstrating the model's potential and its current limitations, while also highlighting the importance of user feedback for continuous improvement.


  • 🚀 Runway ML's Gen 3 is a significant advancement in AI, marking the transition to a new era of AI video generation.
  • 🔍 The video showcases the evolution of AI video from Gen 2 to Gen 3, demonstrating the rapid progress in a short time.
  • 📝 Gen 3 allows for more descriptive prompting, moving away from keyword spamming to a more narrative style.
  • 🎨 Prompt structuring and additional details can greatly improve the quality of the generated video content.
  • 🔑 Including keywords related to subject, action, setting, shot, and style can help in maximizing the generation of desired video outcomes.
  • 💡 It's beneficial to experiment with different orders and combinations of prompt elements to find the most effective prompts.
  • 🌟 The video emphasizes the importance of learning from the community and adapting successful prompts to one's own needs.
  • 🔄 Gen 3's adherence to the prompt can sometimes result in cuts or dissolves if it cannot fulfill a specific request.
  • 🎬 Reusing a seed from a liked generation and making minor adjustments can help maintain a consistent style across iterations.
  • 📚 The transcript mentions a free PDF with shot terms and prompts for Gen 3, available for download on Gumroad.
  • 🚫 Gen 3 has some content restrictions, as seen with certain keywords triggering errors, but workarounds can be found.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is an in-depth exploration of Runway ML's Gen 3, a significant advancement in AI video generation, and providing an ultimate prompting guide for its use.

  • What was the initial capability of Gen 2 when it was first introduced?

    -Gen 2 was initially capable of text-to-video, stringing scenes together to create a narrative, as showcased in the video from April 26th, 2020.

  • How has the prompting style evolved from Gen 2 to Gen 3?

    -The prompting style in Gen 3 has evolved to be more descriptive, allowing users to provide more detail in their prompts rather than focusing on spamming keywords.

  • What is an example of a prompt that was used to test Gen 3's capabilities?

    -An example prompt used was 'the man in Black fled across the desert and the Gunslinger followed,' which resulted in a generated video with some morphing issues.

  • What are the four key elements suggested for an effective prompt in Gen 3?

    -The four key elements suggested for an effective prompt are subject, action, setting, and shot, with additional emphasis on including mood characteristics and style.

  • What is the significance of using keywords associated with these sections in a prompt?

    -Using keywords associated with subject, action, setting, and shot helps to maximize the generation capabilities of Gen 3, ensuring that the AI understands and focuses on the important aspects of the prompt.

  • How can one iterate and improve upon a generated video that they like?

    -One can iterate and improve upon a generated video by reusing the original seed and making adjustments to the prompt, then regenerating the video while maintaining the overall stylistic look.

  • What is the role of the community in learning and improving the capabilities of Gen 3?

    -The community plays a crucial role by sharing their findings, favorite prompts, and techniques, which helps everyone learn what the model is capable of and how to get the best results.

  • What is the importance of rating outputs in Gen 3 Alpha?

    -Rating outputs is important because Gen 3 is still in the alpha phase, and the model will continue to improve based on user feedback and interaction data.

  • What are some of the upcoming features expected in Gen 3?

    -Some of the upcoming features expected in Gen 3 include image-to-video capabilities and possibly a motion brush, although the latter is speculative at this point.

  • Why is it challenging to recreate certain scenes or styles with Gen 3?

    -It can be challenging to recreate certain scenes or styles because Gen 3 tries to adhere strictly to the prompt, and if it can't fulfill a request, it may introduce cuts or dissolves to attempt to meet the prompt's requirements.



🚀 Launch of Runway ML Gen 3: The New Era of AI Video Prompting

The script introduces the third generation of Runway ML, highlighting its advancements over the popular Gen 2 model and marking a significant step into the AI 2.0 era. The narrator discusses an 'ultimate prompting guide' for Gen 3, sharing insights from extensive research and testing. A comparison is made to a previous Gen 2 video from April 20203, showcasing the dramatic improvements in AI video capabilities. The summary also touches on the modern style of prompting, which allows for more descriptive instructions rather than focusing on keyword spamming. Examples of prompts and their outcomes are given, including issues and improvements with additional details and structuring. The importance of learning and adapting to the model's capabilities is emphasized, with a shout-out to other creators for their contributions and a mention of a free PDF resource for further prompts and terms.


🎨 Exploring Advanced Prompting Techniques and Community Insights

This paragraph delves into the intricacies of advanced prompting with Gen 3, discussing the AI's adherence to prompts and its fallback to cuts or dissolves when it cannot fulfill a request. The narrator shares a trick for iterating on a generation by reusing seeds and adjusting settings to maintain a stylistic consistency. Community ideas are explored, such as using the word 'suddenly' for dramatic effect, and the script describes experiments with text and comic book styles, including a Marvel-like opening created by a community member. The narrator also attempts to recreate the Game of Thrones intro using a base prompt and discusses the limitations and workarounds for prompts involving copyrighted content or specific names that trigger content systems.


🌌 Experimenting with Time Lapses and Rating Outputs for Model Improvement

The final paragraph focuses on the AI's ability to handle time lapses and the importance of user feedback for improving the Gen 3 model. The narrator shares an example of a time lapse prompt that successfully created the desired effect and emphasizes the model's current alpha status, encouraging users to rate outputs to aid in its development. There is speculation about future features like image-to-video capabilities and the potential integration of motion brushes, likened to a previously discussed tool. The script concludes with an invitation for users to share their findings and favorite prompts, and a sign-off from the narrator, Tim.



💡Gen 3

Gen 3 refers to the third generation of a product or model, in this case, the successor to Runway ML's Gen 2 model. It symbolizes a significant advancement in AI technology, marking the transition to a new era of AI capabilities. In the video, Gen 3 is presented as having improved prompting capabilities, allowing for more descriptive and less keyword-focused instructions to generate AI videos.

💡Prompting Guide

A prompting guide is a set of instructions or best practices for effectively using a system that requires user input to generate content. In the context of the video, the guide is meant to help users understand how to interact with Gen 3 to create AI videos, emphasizing the importance of descriptive prompts over keyword spamming.

💡Descriptive Prompting

Descriptive prompting involves providing detailed and narrative-like instructions to an AI system to guide its output. The video discusses how Gen 3 allows for this style of prompting, which leads to more nuanced and contextually appropriate video generation compared to the older, keyword-focused approach.

💡Morphing Issues

Morphing issues refer to the visual anomalies that can occur in AI-generated videos where elements or characters change form unexpectedly. The script mentions an example where the 'man in Black' suddenly has an umbrella, illustrating the challenges in AI video generation where the system might not perfectly interpret the prompt.

💡Prompt Structuring

Prompt structuring is the organization of the information within a prompt to optimize the AI's understanding and generation of content. The video suggests that structuring prompts with additional details can lead to improved video outputs, such as specifying the shot type, subject, and other scene elements.

💡Shot Terms

Shot terms are specific types of camera angles or framing used in video production, such as 'wide angle,' 'close-up,' and 'long shot.' The script mentions that Gen 3 recognizes these terms, allowing users to influence the visual style of the generated video by including them in the prompt.


In the context of video generation, 'style' refers to the aesthetic or visual approach taken in the presentation of the content. The video discusses how using style keywords, such as 'cinematic' or 'IMAX,' can influence the final look of the AI-generated video, reinforcing the desired atmosphere or mood.


In AI video generation, a 'seed' is a value that helps initialize the randomness in the model's output generation process. The script explains how users can reuse a seed from a previous generation to maintain stylistic consistency when iterating on a prompt.

💡Time Lapses

Time lapses are a video technique where time is condensed, showing hours or days passing in a short video. The video mentions that Gen 3 does a good job with time lapses, especially when they involve different intervals, as demonstrated in the prompt about a woman sitting by a window with rapidly changing day to night.

💡Rating Outputs

Rating outputs involves providing feedback on the AI-generated content, which helps the system learn and improve. The script encourages users to rate their outputs, noting that Gen 3 is still in the alpha phase and relies on user feedback for refinement.

💡Image to Video

Image to video refers to the capability of an AI system to create a video sequence from a single image. The script hints at upcoming features in Gen 3 that will include this functionality, expanding the range of creative possibilities for video generation.


Introduction to Runway ML's Gen 3, a major advancement in AI video generation.

Comparison of Gen 2 and Gen 3, showcasing the evolution from text-to-video to a more advanced model.

The new prompting style of Gen 3 allows for more descriptive prompts rather than keyword spamming.

Example of a Gen 3 prompt resulting in a scene with morphing issues.

Improvement in Gen 3's output by adding details and structuring the prompt.

Importance of learning and adapting to what the new model is capable of, including borrowing successful prompts from others.

The significance of including keywords associated with subject, action, setting, and shot in your prompts.

Availability of a free PDF with shot terms and prompts for Gen 3 experimentation.

The flexibility in prompt order and the encouragement to experiment with different structures.

The impact of style keywords like 'cinematic' and 'IMAX' on the output of Gen 3.

Gen 3's adherence to prompts, even to the point of inserting cuts or dissolves to match descriptions.

Technique of reusing a seed from a liked generation to maintain a stylistic consistency.

Community ideas and experiments with the word 'suddenly' in prompts for interesting effects.

Examples of Gen 3's capabilities with text, including mimicking the MCU opening.

Potential issues with content systems when using certain keywords and how to navigate them.

Creative use of prompts to generate unique AI videos, such as a miniature civilization on an ancient scroll.

The limitations of Gen 3 in creating videos from actual script pages, compared to the Dream Factory video.

The effectiveness of Gen 3 with time-lapse prompts and its potential applications.

The importance of rating outputs to help improve the Gen 3 model during its alpha phase.

Anticipation for future features in Gen 3, such as image-to-video capabilities.

Invitation for the community to share findings and favorite prompts for collaborative exploration.