Get Consistent Character and Styles with Dalle-3
TLDRIn this video, the host discusses the use of custom instructions with Dalle-3, a tool for generating images and stories. They were inspired by a video from Jil Gilberry, who shared his custom instructions for Dalle-3. The host applies these instructions to create a comic book-style story featuring a character named Blake, who transforms into the Sentinel. The process involves careful character description and art style definition, with the host emphasizing the importance of being less descriptive initially to allow for more variety in character positioning. Despite some challenges, such as Dalle-3 dropping the character from a few panels, the host is pleased with the outcome. They also highlight the diversity and creativity of the generated images. The video concludes with the host's strategy for refining the generated content using Photoshop and other editing tools, and their plan to use the generated panels for a voice-over project.
Takeaways
- 🎨 Use custom instructions with Dalle-3 to get consistent characters and styles in your artwork.
- 📚 Follow Jil uh Gilberry's YouTube channel for tips on using custom instructions effectively.
- 🔗 Check out the custom instructions link provided for Dalle-3 to see examples of the results you can achieve.
- 🖌 When describing characters, be less descriptive initially to allow for more variety in positioning.
- 🧩 For comic creation, convert the custom instructions to fit the visual medium, as text-based rules need to be adapted for images.
- ✂️ Edit the output as needed to refine the artwork and ensure it meets your vision.
- 📝 Keep a clear description of your main characters, art style, and superhero look as your 'North Star' to guide the creative process.
- 📉 Be flexible with other elements but maintain the core elements that define your project's identity.
- 📖 Give the AI a simple prompt and direction to create a focused and relevant story.
- 🚫 Be cautious with language that could trigger content filters, and adjust your prompts accordingly to avoid issues.
- 🖥️ Use image editing software like Photoshop for fine-tuning the final images to perfection.
Q & A
What was the main topic discussed in the video?
-The main topic discussed in the video was how to use custom instructions with Dalle-3 to create consistent characters and styles in comic art.
Who is Jil uh Gilberry and what does he do?
-Jil uh Gilberry is a content creator with a YouTube channel where he talks about various topics, including custom instructions for AI like Dalle-3, which the speaker found helpful for their work.
What is a custom instruction in the context of Dalle-3?
-A custom instruction in the context of Dalle-3 is a set of guidelines or rules provided to the AI to control the output, such as the style, background, and specific characteristics of the generated images.
Why is it important to be less descriptive when describing characters for the first time using Dalle-3?
-It is important to be less descriptive when describing characters for the first time to allow the AI more variety in how it positions them, as Dalle-3 tends to carry over the initial stance and position of characters in subsequent images.
What is the term used to describe the main characteristics that the speaker wanted to maintain consistency for in the comic?
-The term used is 'North Star,' which refers to the main character, superhero look, and art style that the speaker wanted to maintain consistency for in the comic.
What was the process for creating a story with Dalle-3?
-The process involved giving Dalle-3 a simple prompt, specifying the main characters, art style, and a direction for a two-minute story. The AI then generated panels in the style of an action comic book, which were later refined and edited by the speaker.
What issue did the speaker encounter when generating the panels?
-The speaker encountered an issue where Dalle-3 dropped the main character from some of the panels and included imagery that was not part of the custom instructions, such as soldiers grabbing a woman, which was against the intended content.
How did the speaker address the issue of Dalle-3 not generating the desired character in some panels?
-The speaker re-generated the image using the character description as a 'North Star' and ensured that the custom instructions were followed, which eventually led to the generation of the desired character.
What was the final step the speaker took to complete the comic?
-The final step was to instruct Dalle-3 to take the panels and convert them into a narration, which could then be used for voice-over work in combination with the generated images.
What percentage of completion did the speaker consider the final comic to be?
-The speaker considered the final comic to be about 80% complete and decided to stop there, planning to use 11 labs to finalize the project.
What advice does the speaker give for achieving perfect results with Dalle-3?
-The speaker suggests that for perfect results, it's best to work with an illustrator, Photoshop, or other image editing software to fine-tune the generated images according to one's requirements.
Outlines
🎨 Custom Instructions and AI Art in Dolly 3
The speaker begins by welcoming viewers back to their show, where they discuss the use of variables to create consistent characters in Dolly 3, an AI art tool. They mention a recent live session and their desire to refine their approach after watching a video by Jil (possibly a name error, should be Gil) berry, who has a YouTube channel. The speaker is inspired by the concept of custom instructions, which they have used extensively in their work. They describe how custom instructions can guide the AI, similar to providing rules for output. They note that they needed to convert the instructions for use with comics, as Dolly 3 is more focused on images and photography. The speaker shares their process of using custom instructions to generate a comic, including the importance of being less descriptive when positioning characters to allow for more variety. They also mention the editing process and how they arrived at a satisfactory result, emphasizing the 'North Star' elements that must be present in their AI art: the main character, superhero look, and art style.
📚 Crafting a Story with Custom Instructions
The speaker outlines their process of creating a story using Dolly 3, starting with a simple prompt to generate a two-minute story about a character named Blake in a coffee shop who then transforms into the Sentinel upon encountering an army in dark orange. They emphasize the importance of direction in the prompt to avoid generic or tangential stories. The AI generates a story in the form of comic book panels, which the speaker appreciates for its structure and style. However, they encounter issues with Dolly 3 when it generates images that do not include their main character, possibly due to the use of certain descriptive language triggering the AI's content rules. The speaker discusses the need to adjust the language in the prompt to avoid such issues and highlights the importance of their 'North Star' elements in maintaining the integrity of the character and story. They also mention their satisfaction with certain panels and their plan to fix minor issues in post-production using Photoshop.
🖌️ Diverse Representation and Final Touches
The speaker highlights the diversity in the generated images, noting the variety of expressions and the inclusion of different faces in the artwork. They discuss the creative and imaginative aspects of Dolly 3, which sometimes leads to unexpected results. The speaker iterates their process of toning down certain elements of the story to avoid triggering content rules and maintaining the focus on their main character, Blake, who transforms into the Sentinel. They appreciate the camera angles and the quality of the images generated, despite some inconsistencies with the character's attire. The speaker concludes by mentioning their final step of converting the panels into a narration for voice-over work, rounding out their creative process. They acknowledge the process is not perfect and suggest using additional software for fine-tuning the images, but decide to consider their work about 80% complete, indicating a high level of satisfaction with the results.
Mindmap
Keywords
💡Dalle-3
💡Variables
💡Custom Instructions
💡Comic Conversion
💡Art Style
💡Character Description
💡North Star
💡Action Comic Book
💡Panels
💡Editing
💡Diversity
💡Illustrator and Photoshop
Highlights
The speaker discusses the use of variables to achieve consistent characters in Dolly 3.
Inspired by a video by Jil gilberry, the speaker explores custom instructions for Dolly 3.
Custom instructions allow for more control over the output, similar to setting rules for output.
The speaker needed to convert custom instructions for comic images, which Dolly 3 primarily handles.
Editing was required after the conversion to better suit the desired comic format.
When describing characters in Dolly 3, it's advised to be less descriptive for more variety in positioning.
The speaker was more descriptive with the Sentinel's suit to achieve a specific look.
The art style described was 'Western modern comic style', which influenced the output.
Custom instructions were used to maintain the North Star elements: main character, superhero look, and art style.
The speaker provided a simple prompt to Dolly 3 to generate a two-minute story in an action comic style.
Dolly 3 produced a story in panels, which was unexpected and appreciated by the speaker.
The speaker encountered issues with Dolly 3 generating inappropriate content, which was manually corrected.
Jil Patrice's prompt specified avoiding PG-13 language to align with Dolly 3's content rules.
The speaker emphasizes the importance of the North Star elements in maintaining consistency throughout the process.
Dolly 3's output was imaginative and sometimes created its own prompts, which could be a double-edged sword.
The speaker had to adjust the language used in the prompt to avoid triggering Dolly 3's content restrictions.
Despite some manual adjustments, the speaker was impressed with the final output and the diversity presented in the images.
The speaker suggests using Photoshop or similar software for fine-tuning the generated images.
The final step was to convert the panels into a narration for voice-over work.
The speaker concludes by stating the process is about 80% complete and encourages continued creativity.