AI Text Styles
Make your words pop with AI text styles
Overview
In the realm of digital creativity, the introduction of AI Text Styles marked a revolutionary step for Picsart. While many of the GenAI tools out there don’t serve too many actual use cases yet, this feature emerged to break the boundaries of conventional text formatting, offering a unique blend of visual appeal and technological innovation.
Defining the Problem
Before AI Text Styles, text customization in Picsart was limited and lacked the dynamic flair users craved. The challenge was to infuse creativity and versatility into text styling, elevating user engagement and expression.
Furthermore, with conventional text styling and formatting tools available across the competition, we were determined to create a novelty tool that would ensure the competitive advantage.
User Research and Hypotheses
Research revealed a strong user desire for personalized, eye-catching text styles. The hypothesis was that AI Text Styles would attract a diverse user base, enhance creative expression, and simplify complex design tasks.
My Roles
In this transformative journey to elevate Picsart’s mobile video editor, I wore multiple hats, each contributing to a different facet of the project.
Leading end-to-end Design
I led the project through every design phase, from initial problem framing to final quality checks. This involved user research, competitor analysis, ideation, prototyping, and ongoing QA to ensure our solutions were user-centric and business-aligned.
Research Lead
I spearheaded our user research initiatives, employing a mix of qualitative and quantitative methods to gain valuable insights. This data-driven approach informed our design decisions and helped us better understand our users’ needs and behaviors.
Interactive Prototyping
I developed dynamic prototypes to test design concepts and identify usability issues. These prototypes also simplified the handover to developers by illustrating the feature’s intended behavior.
Documentation
I meticulously recorded each step of our design journey, including the rationale behind our decisions and why certain variants were chosen over others. This comprehensive guide served as a blueprint for the team, ensuring alignment and facilitating smooth project execution.
Co-defining the PRD
Working closely with product managers, I helped shape the Product Requirements Document. Our collaborative approach was crucial, as some product requirements, element relations, and interaction details only emerged during the design process. This made the PRD a living blueprint, aligning with design and business goals.
Design QA
As the final gatekeeper, I ensured that the developed features matched the original designs. My QA checks were the last line of defense, confirming that our user experience met both design and functional expectations.
In all these roles, my focus was on contributing to a user-centric design that would meet both user needs and business objectives, while maintaining active collaboration with all stakeholders to ensure consistency, sustainability, and that everyone was on the same page.
The user personas
- Creative Enthusiasts: Seeking innovative ways to make their content stand out.
- Social Media Influencers: Looking for quick, unique text styles to boost engagement.
- Small Business Owners: Needing easy-to-use tools for creating professional, branded content.
Ava Chen
🙋♀️ 29 years old
👩🎓 BA in Graphic Design
👧 Single
🌐 Freelance Graphic Designer
🇺🇸 San Francisco
 need to quickly create visually striking designs for my clients.
 🎯 Goals
- Efficiently produce unique, attention-grabbing graphics
- Utilize innovative text styles to enhance designs
- Balance multiple client projects with diverse style needs
😤 Frustrations
- Time-consuming design processes
- Limited text styling options in standard design tools
- Keeping up with current design trends
📒Â Story
Ava is a freelance graphic designer in San Francisco, known for her quick turnaround and creative designs. She constantly seeks tools like the AI text styles in Picsart to enhance her projects, ensuring they stand out in the competitive freelance market.
Liam Johnson
🙋♂️ 34 years old
👨🎓 Master’s in Marketing
🤵 Married
🌐 Social Media Marketing Consultant
🇬🇧 London
Every social media post must be a visual masterpiece to capture our audience.
 🎯 Goals
- Create visually compelling social media content
- Quickly adapt to different brand aesthetics for various clients
- Stay ahead of social media trends
😤 Frustrations
- Limited resources to create fresh, unique content
- Balancing creativity with quick content production
- Keeping content visually appealing and on-brand
📒Â Story
Liam is a social media marketing consultant in London, working with several brands to enhance their online presence. He relies on tools like AI text styles to quickly produce visually distinct and brand-aligned content for various social media platforms.
Technical and Creative Challenges
Creative Challenges
Developing AI Text Styles involved overcoming significant creative challenges, particularly in working with technologies like Stable Diffusion and ControlNet. Creating high-quality, consistent, and clean styles for characters was crucial. This task became a pivotal make-or-break aspect of the project. Through extensive experimentation with numerous character masks and AI model configurations, key discoveries were made:
- Importance of Font Choice: The font used in the character mask significantly influences the quality of the final output.
- Style-Specific Performance: Certain styles, like “Balloon,” “Ice,” and “Pizza,” yielded more realistic results than others, such as “Spaghetti.”
These insights led to the selection of five initial styles for the MVP release, ensuring high-quality and ultra-realistic results.
Technical Challenges
The project’s technical side also posed significant challenges. The use of Text-to-Image AI models, which produce raster images, introduced several user experience issues:
- Destructive Conversion: Converting text layers to image layers was a destructive process, hindering further editing.
- Toolset Change: The unintentional layer conversion created UX friction due to the change in available toolsets.
- Limited Text Formatting Options: Once converted to an image, further text tools and formatting options became unavailable.
- Inability to Edit Text: Users lost the ability to edit text post-conversion.
- Style Selection Limitation: Converting to an image layer prevented users from switching styles without starting over.
To address these issues, the team collaborated closely with AI and backend engineers, exploring various solutions. The breakthrough came with the idea of compiling separately generated characters into font packages. This innovative approach resolved the initial problems, enabling the use of additional text tools like “Bend,” “Letter Spacing,” and “Line Height” in conjunction with AI Text Styles. This solution not only enhanced the user experience but also laid the groundwork for future improvements and expansions of the feature.
AI Text Styles
Make your words pop with AI text styles
In the realm of digital creativity, the introduction of AI Text Styles marked a revolutionary step for Picsart. While many of the GenAI tools out there don’t serve too many actual use cases yet, this feature emerged to break the boundaries of conventional text formatting, offering a unique blend of visual appeal and technological innovation.
Note: Unfortunately, the most exciting details can’t be disclosed [yet] due to confidentiality.
Impact on post-launch
The AI Text Styles tool performed very well on the MVP release.
- The tool’s open > export rate is 41.24%
- 86.24% of users who opened the tool tried it
- The gold users are significantly converting better in the funnel
Hypothesis
The initial hypothesis based on user research for prioritizing AI Text Styles proved to be correct as the post launch stats are showing. The open > export rate is very promising and it suggests that it can be further improved by introducing more styles, more free options, more fonts and improved UX in general.
The gold users are converting significantly better which suggests introducing more free presets would help with improving the conversion rates.
Next Steps and Future Improvements
As the engineering behind AI Text Styles continues to evolve, we are looking forward to expanding the feature’s capabilities significantly. Future enhancements include:
- Style Prompting: Alongside the preset styles, we plan to introduce style prompting, allowing users to have more control and input in the style generation process.
- Multiple Font Choices: We aim to offer a broader range of font selections, giving users greater flexibility in their creative endeavors.
- Introducing more free presets: The free users convert less than gold users. Providing them with more free presets will improve their conversion and retention rates.
- Feedback Loop Integration: Implementing a feedback loop will be a key focus. This will enable user interactions to contribute directly to the AI model’s data flywheel, enhancing the model’s performance and accuracy over time.
These steps are geared towards making AI Text Styles not just a tool for creating engaging visuals, but also an evolving platform that adapts and improves based on user input and technological advancements.
Conclusion
The AI Text Styles project for Picsart represents a significant stride in digital creativity, blending innovative AI technology with user-centric design. Overcoming creative and technical challenges, the team developed a feature that not only meets but surpasses user expectations in text styling. The introduction of style prompting, a wider array of font choices, and a feedback loop marks the next phase in the feature’s evolution. These improvements are not just enhancements; they symbolize a commitment to continual learning and adaptation, leveraging user feedback to refine and perfect the technology. As the project advances, AI Text Styles is poised to redefine the landscape of digital text customization, offering users unparalleled creative freedom and expression.
Try AI Text Styles for yourself!
You can try it on web and the iOS app now.