This New AI Tool Does Consistent Characters (For Real)

Theoretically Media
4 Jan 202413:36

TLDRIn this video, Tim introduces 'Scenario,' an AI image generator that excels at creating consistent characters, a feature that has been challenging to achieve in AI imagery. He discusses the free starter plan and its generous features, guiding viewers through the process of creating a custom character. Tim explains how to blend models, use prompts effectively, and iterate to refine the character's appearance from different viewpoints. He also highlights the importance of negative prompts and demonstrates the use of 'Control Net' for specific poses. The video showcases the potential of Scenario for character creation and customization, and Tim's enthusiasm for the platform's future development is evident.


  • ๐ŸŽจ **Scenario AI** is a new image generator that can create consistent characters in AI imagery, which has been a challenge until now.
  • ๐Ÿš€ The tool has its origins in creating AI assets for games, offering a unique approach to character generation.
  • ๐Ÿ†“ **Scenario** offers a free starter plan that is quite generous, allowing for extensive testing and usage without cost.
  • ๐ŸŒŸ The process involves selecting a model, generating an image, and then fine-tuning it with sliders to blend different styles.
  • ๐Ÿ“ There's a prompt box with features like prompt builder and prompt spark to assist in generating ideas and refining the output.
  • ๐Ÿšซ Negative prompts are crucial in Scenario to guide the AI away from undesired elements in the generated images.
  • ๐Ÿงฐ Users can start with a template for character creation or upload their own, then refine it through an iterative process.
  • ๐Ÿ”„ Iteration is key, as the AI will make odd choices, and it may take several attempts to achieve a coherent character across different views.
  • ๐Ÿ–ผ๏ธ Once a character is created, it can be upscaled and edited in a photo editor like Photoshop for further refinement.
  • ๐Ÿค– Training your own model with various views of the character helps the AI learn and generate more consistent results.
  • ๐Ÿ”„ After training, blending the model with others can introduce new styles while maintaining the character's consistency.
  • โ›“ Control net is a feature that allows for specific poses and more control over the final image, although it may require some trial and error.

Q & A

  • What is the main challenge in creating consistent characters in AI imagery?

    -The main challenge has been achieving consistent character representation across different AI-generated images, with results often being inconsistent or 'janky'.

  • What is the name of the AI image generator discussed in the transcript?

    -The AI image generator discussed is called 'Scenario'.

  • What is the basis of the Scenario AI image generator's development?

    -Scenario was initially developed for creating AI assets for games.

  • How does the Scenario platform handle the creation of a custom character?

    -Scenario allows users to select a model, generate an initial image, and then refine it using prompts and sliders to control the influence of different aesthetic styles.

  • What are the features of the Scenario platform that assist in character creation?

    -Scenario features a prompt box with a prompt builder for keyword suggestions, a prompt spark for generating ideas based on a basic prompt, and a reference image section for character creation.

  • What is the significance of negative prompts in Scenario?

    -Negative prompts play a crucial role in refining the output by specifying what should not be included in the generated images.

  • How does one begin training a custom model in Scenario?

    -After preparing the character images, users go to the models section, select 'Train your own model', and upload the character images while providing descriptions for each view.

  • What is the purpose of blending models in Scenario?

    -Blending models allows users to combine the characteristics of different models to achieve a desired style or to add more detail and complexity to the generated images.

  • How can users ensure their custom character remains consistent across different images?

    -Users can ensure consistency by using the training feature to teach the AI model to recognize the character from different viewpoints and by using blending and control net features to refine poses and styles.

  • What is the 'canvas' function in Scenario used for?

    -The canvas function allows users to edit generated images, such as removing unwanted elements or making adjustments to the final character representation.

  • What additional tool was mentioned for enhancing the upscaled character images?

    -The transcript mentions using 'Magnific' for a creative upscale, which has been a go-to tool for enhancing image quality.

  • How does Scenario's layers tab assist in the character creation process?

    -The layers tab in Scenario allows users to try out different generations of their character, providing more control and options for refining the final character design.



๐ŸŽจ Introduction to Consistent Character AI Imagery

The video begins by addressing the challenge of creating consistent characters in AI imagery, a task often referred to as a 'Holy Grail'. The speaker shares their previous experience with AI image generators and introduces Scenario, a new AI image generator with origins in game development. The video promises to explore a solution that many viewers have been seeking. The workflow was discovered through a thread by Vlados AI, a customer success specialist and AI artist at Scenario. The platform offers a free starter plan, which is quite generous, and the video's testing was primarily done using this plan before upgrading to the essential plan for final touches. The process involves selecting a model, generating an image, and using sliders to blend different aesthetic styles. The speaker also discusses the importance of negative prompts and the use of a reference image to begin character creation.


๐Ÿ–ผ๏ธ Character Creation and Model Training in Scenario

The speaker details the process of creating a custom character in Scenario by starting with a template and adjusting the influence of different models to achieve the desired aesthetic. They mention the use of an upscale button for image enhancement and suggest using external tools like Magnific for further upscaling. The character is then prepared by separating different views (front, side, and back) and adding them to the model for training. The training process involves naming the character and providing descriptions for each view. The speaker cautions against immediately using a newly trained model due to the risk of bland results and suggests blending the model with another to improve the outcome. They also touch on the importance of negative prompts in refining the character's appearance and the option to use image referencing or control net for specific poses.


๐Ÿ” Refining Character Poses and Final Thoughts on Scenario

The video continues with techniques for achieving specific character poses, either through image referencing or using control net, which offers more control over the character's pose. The speaker demonstrates how to use control net to adjust the character's pose and mentions the need for multiple iterations due to the unpredictable nature of AI art generation. They highlight the importance of negative prompts, particularly for avoiding unwanted features like elf ears. The speaker also discusses the use of Scenario's canvas function for editing out unwanted elements in the generated images. They conclude by expressing excitement about the potential of Scenario as a young platform and encourage viewers to explore it themselves, providing a link to a PDF guide for creating consistent characters and thanking the audience for watching.



๐Ÿ’กConsistent Characters

Consistent characters refer to the ability to generate images of the same character with a consistent appearance across different contexts or poses. In the video, this is the primary goal and challenge that the AI tool aims to address, as it has been historically difficult to achieve in AI imagery.

๐Ÿ’กAI Image Generator

An AI image generator is a software or tool that uses artificial intelligence to create images based on given prompts or descriptions. In the context of the video, the AI image generator is the core technology that the host is exploring and discussing.


Scenario is an AI image generator mentioned in the video that has its roots in creating AI assets for games. It is highlighted as a solution for generating consistent characters in AI imagery, which is a significant advancement in the field.

๐Ÿ’กFree Starter Plan

The free starter plan is a tier of service offered by Scenario that allows users to utilize the AI image generator for free, with certain limitations compared to paid tiers. The video mentions that the free tier is quite generous and sufficient for the testing conducted.

๐Ÿ’กModels Tab

The models tab in Scenario is a feature that allows users to select different aesthetic styles for their AI-generated images. It is part of the process of creating custom characters and is one of the first steps mentioned in the video.

๐Ÿ’กInfluence Sliders

Influence sliders are tools within the Scenario interface that control the degree to which different models affect the final output image. They are used to blend styles, such as 'Illustrated anime' and 'creepy Nordic fairy tale', to achieve a desired look.

๐Ÿ’กPrompt Box

The prompt box is a user input field in Scenario where users can type in descriptions to guide the AI in generating the desired image. It includes features like prompt builder and prompt spark to assist users in crafting effective prompts.

๐Ÿ’กNegative Prompts

Negative prompts are user-specified elements that the AI should avoid including in the generated image. They are important in Scenario for fine-tuning the output and ensuring that the AI does not include unwanted features, such as 'elf ears' in the video example.

๐Ÿ’กImage to Image

Image to image is a mode in Scenario where the AI uses an existing image as a reference to generate a new image. It is used in the character creation process to maintain consistency across different views of the character.

๐Ÿ’กTraining Your Own Model

Training your own model in Scenario involves providing the AI with multiple images of a character from different angles or views. The AI then 'learns' to recognize and replicate the character consistently, which is a key aspect of creating consistent characters.

๐Ÿ’กControl Net

Control net is a feature in Scenario that allows for more precise control over the pose and positioning of elements in the generated image. It is used to achieve specific poses for characters, making the generated images more aligned with the user's vision.

๐Ÿ’กCanvas Function

The canvas function in Scenario is a tool that enables users to edit the generated images directly within the platform. It is mentioned in the context of fixing issues like 'two Siri heads' in an image, providing a way to refine the output without needing external photo editing software.


AI imagery has long been a challenge in creating consistent characters, often resulting in 'janky' outcomes.

Scenario is a new AI image generator with roots in creating AI assets for games.

The platform offers a free starter plan that is quite generous for testing and creating characters.

Users can blend and remix different aesthetic styles to create unique character models.

A prompt builder and prompt spark feature help generate ideas and refine searches.

Negative prompts play a significant role in shaping the final output in Scenario.

Character creation begins with a template and iterative prompting for different character views.

The influence of each model can be adjusted using sliders to control the output style.

Upscaling options are available for enhancing the character image quality.

Photoshop or similar editors can be used to separate character views for further refinement.

Training your own model with various character views helps in creating a consistent character.

Advanced settings should be used with caution to avoid overly bland or incorrect results.

Blending your model with another can add depth and variety to the character's appearance.

Negative prompting can be used to correct unwanted features, such as elf ears in a fantasy setting.

Control net allows for specific poses and detailed control over the character's depiction.

Even with AI art, multiple iterations are often needed to achieve the desired outcome.

Scenario's canvas function offers basic editing capabilities to correct minor issues in the generated images.

The platform provides a layers tab for trying out different generations of a character.

Scenario is a promising tool for creating consistent characters in AI imagery, with much potential for future development.