purple
Glad to see you
Mobile is under construction.
Please refer to desktop
purple
Glad to see you
Mobile is under construction.
Please refer to desktop
The Project Tags
Desktop
Desktop
SaaS
AI Technology
Design Software
AI Technology
Design Software
In the beginning
As creatives ourselves, we noticed that building with generative AI often meant jumping between multiple tools. Creating images, editing assets, and refining prompts usually required several different platforms and subscriptions, forcing users to manage their creative process across multiple environments. My co-founders and I saw an opportunity to simplify that experience by exploring what it might look like to generate and manage AI assets from a single visual workspace. That became the foundation for Purple, with a longer-term vision of optimizing the entire creative AI workflow from the moment a user has an idea, to writing prompts, generating assets, and shaping everything in between.
Purple today....
Purple is an experimental AI creation tool that allows users to generate digital assets by visually connecting prompts and AI models through nodes. The product is currently live and evolving as we explore new ways to make AI workflows easier to build and understand.
Access a Model Library: Choose from a library of generation models directly inside Purple. Users can quickly switch between models to experiment with different outputs without leaving their workflow.
Node-Based Prompting: Build AI workflows by visually connecting prompt blocks to model blocks. Nodes allow you to structure how ideas move from prompt to generation in a flexible, easy-to-follow way.
Open Canvas: Work across all of your generations in one visual workspace. The open canvas lets you organize prompts, models, and outputs in a single place so your workflow stays visible as it grows.
But what about tomorrow..?
While building Purple and getting feedback from users we uncovered additional areas of improvement for the overall value we could add to user’s workflows.
We conducted five persona interviews with creatives to better understand how they currently use generative tools and where their workflows break down. We uncovered some exciting insights and bucketed them into a few key themes, which we tackled in an internal hackathon to explore possible product directions before committing to the roadmap.
From Insights To Explorations
From Insights To Explorations
Node-based systems become harder to follow as workflows grow in size or used less often.
Node-based systems become harder to follow as workflows grow in size or used less often.
We experimented with ways to make our node workflows easier to read by labeling prompts within a prompt concatenator. We also studied how other tools approach AI workflow visualization to identify patterns that could improve clarity inside Purple.
We experimented with ways to make our node workflows easier to read by labeling prompts within a prompt concatenator. We also studied how other tools approach AI workflow visualization to identify patterns that could improve clarity inside Purple.
Workflow Visualization
Workflow Visualization
Experiment
Experiment
Insight
Insight
Users’s lack of knowledge on how to structure and edit their prompts results in unnecessary generations.
Users’s lack of knowledge on how to structure and edit their prompts results in unnecessary generations.
We explored a hybrid drop down open ended prompt editor that could make prompts easier to understand and refine. The idea was to help users see which parts of their prompt were influencing the output, we are conducting more exploration on how this could be done without making the experience feel too restrictive.
We explored a hybrid drop down open ended prompt editor that could make prompts easier to understand and refine. The idea was to help users see which parts of their prompt were influencing the output, we are conducting more exploration on how this could be done without making the experience feel too restrictive.
Prompt Optimization
Prompt Optimization
Experiment
Experiment
Insight
Insight
Users want to know which AI tool to use to generate their desired asset using the least amount of tokens possible.
Users want to know which AI tool to use to generate their desired asset using the least amount of tokens possible.
We prototyped an enhanced model library window to include more information about the specific AI model. Our hypothesis here is that the more information you have on a model, the better you can decide when and where to use it.
We prototyped an enhanced model library window to include more information about the specific AI model. Our hypothesis here is that the more information you have on a model, the better you can decide when and where to use it.
Token Efficiency
Token Efficiency
Experiment
Experiment
Insight
Insight
What we’re exploring next
These early prototypes helped us test possible directions for improving the AI workflow experience. While Purple currently functions as a node-based generation tool, our research highlighted a broader opportunity to make complex workflows easier to understand and manage. Moving forward, we are exploring how Purple can better support prompt optimization and workflow visualization as creators build increasingly complex AI pipelines.