What if you don't know the name of the color you're looking for?
Choosing the right color can be difficult, especially for people with limited experience in color theory. Most digital color tools rely on hex codes, RGB sliders, and color wheels, requiring users to translate the color they imagine into a technical format. For novice designers, this can involve trial and error.
At the same time, people often describe colors differently in everyday life. We say:
rather than starting with an RGB value or hex code. That became the starting point for this proposal:
Could color-selection tools build on the way people naturally describe and recognize colors, while also helping them become more confident with color over time?
Previous research suggests that people with limited color-theory knowledge can have difficulty selecting specific colors using conventional digital tools. Other work has explored richer ways of interacting with color, including mixing, comparing, and manipulating palettes. However, these approaches do not necessarily reflect the way people describe colors in everyday language.
Research on color naming and memory colors suggests that people frequently connect colors to familiar objects and experiences, such as dry grass, green grass, red brick, and beach sand.
While existing color tools are good at representing color technically, they may not always support the way people naturally describe and learn color.
To address this gap, I proposed a system with two complementary components.
01 Natural Language Color Picker
Users could describe the color they are imagining through an object, phrase, or real-world association instead of beginning with a color wheel or technical value.
Rather than mapping a phrase to one predefined color, the proposed system would give users multiple interpretations to explore. For example, “summer ocean” could produce a group of colors extracted from relevant real-world images.
Enters “summer ocean”
Retrieves real-world images associated with the description
Extracts representative colors from those images
Ranks and displays common color associations
Learning is also an important goal. Users could double-click a color to reveal its RGB and hex breakdown,
Try it: Double-click the swatch!
Try it: Double-tap the swatch!
while a floating hex tooltip would appear as they draw, connecting natural-language descriptions with technical color representations.
Try it: Hover over the colored area!
Try it: Tap the circle!
By connecting something familiar to its technical representation, the tool could help users gradually develop familiarity with RGB and hex values while selecting colors.
02 AI-Assisted Palette Generator
The system analyzes what the user is designing and generates color palettes based on relevant real-world imagery.
Novice designers may also struggle with deciding which colors work together or how a palette could apply across an entire design.
The proposed AI-assisted Palette Generator would analyze a user's line sketch by detecting visual cues such as shapes and dominant regions.
Analyzes the sketch
Retrieves contextually similar real-world images
Extracts representative colors from those images
Generates and ranks palette suggestions
Applies selected colors to corresponding regions of the drawing
When a user selects or adjusts a palette color, the corresponding regions of the sketch would update automatically, allowing them to see how the changes affect the overall design.
The system would also log users' final selections to refine future recommendations. When users choose a suggested color swatch or palette as their final selection, that choice would contribute to its ranking, allowing frequently selected options to rank higher for future users.
What happens when color selection becomes more intuitive and contextual?
The proposal explores three research questions:
RQ1: How can natural language input and AI-assisted palette generation support intuitive color selection, and how do these systems influence users' creativity and learning in color theory?
RQ2: How do these systems affect users' engagement and confidence during the color-selection process?
RQ3: How might using natural language in the color-selection process improve usability and accessibility for users with color-vision deficiency?
(The accessibility question would be evaluated only if participants with color-vision deficiency were represented in the study sample.)
This project was developed as an HCI research proposal rather than a completed or evaluated product. I proposed a two-phase comparative study to examine both task performance and the experience of creating with color.
Phase 1 Recreate
Participants would recreate a reference image using either a conventional color picker or the proposed system. I would compare color accuracy, completion time, number of adjustments, and recall of RGB/hex values.
Phase 2 Create
Participants would switch tools and freely color a second image. This phase would examine creativity, exploration, engagement, confidence, and how participants use or modify the suggested palettes.
The next step would be to implement the proposed interactions and evaluate them against conventional digital color-selection tools.