Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LACE: Controlled Image Prompting and Iterative Refinement with GenAI for Professional Visual Art Creators

Yenkai Huang, Ning Zheng

TL;DR

LACE integrates a Latent Auto-recursive Composition Engine into Adobe Photoshop to give professional visual artists fine-grained, layer-based control over AI-generated content. The system supports turn-taking and parallel collaboration to address expressivity, coherence, and workflow compatibility challenges in AI-assisted creation. In a pilot with 21 participants, LACE yielded higher usability, ownership, satisfaction, and artistic perception than text-only baselines, with significant p-values ($p=0.039$, $p=0.009$, $p=0.003$, $p=0.005$). Key findings show that layer-based prompting and flexible collaboration modes enable more concept exploration early and more precise refinement later. The work suggests broader applicability to other creative domains and highlights the importance of user control in AI-assisted workflows.

Abstract

We present LACE, a hybrid Human-AI co-creative system integrated into Adobe Photoshop supporting turn-taking and parallel interaction modes for iterative image generation. Through a study with 21 participants across representational, abstract, and design tasks, we found turn-taking preferred in early stages for idea generation, and parallel modes suited for detailed refinement. While this shorter workshop paper provides key insights and highlights, the comprehensive findings and detailed analysis are presented in a longer version available separately on arXiv.

LACE: Controlled Image Prompting and Iterative Refinement with GenAI for Professional Visual Art Creators

TL;DR

LACE integrates a Latent Auto-recursive Composition Engine into Adobe Photoshop to give professional visual artists fine-grained, layer-based control over AI-generated content. The system supports turn-taking and parallel collaboration to address expressivity, coherence, and workflow compatibility challenges in AI-assisted creation. In a pilot with 21 participants, LACE yielded higher usability, ownership, satisfaction, and artistic perception than text-only baselines, with significant p-values (, , , ). Key findings show that layer-based prompting and flexible collaboration modes enable more concept exploration early and more precise refinement later. The work suggests broader applicability to other creative domains and highlights the importance of user control in AI-assisted workflows.

Abstract

We present LACE, a hybrid Human-AI co-creative system integrated into Adobe Photoshop supporting turn-taking and parallel interaction modes for iterative image generation. Through a study with 21 participants across representational, abstract, and design tasks, we found turn-taking preferred in early stages for idea generation, and parallel modes suited for detailed refinement. While this shorter workshop paper provides key insights and highlights, the comprehensive findings and detailed analysis are presented in a longer version available separately on arXiv.

Paper Structure

This paper contains 11 sections, 9 figures.

Table of Contents

  1. INTRODUCTION
  2. LACE SYSTEM
  3. Layer-Based Prompting
  4. Flexible Collaboration Modes
  5. STUDY AND KEY FINDINGS
  6. Key Findings
  7. Improved Usability and Ownership
  8. Correlation Between Ownership and Usability
  9. Engagement in Structured Tasks
  10. DISCUSSION AND FUTURE WORK
  11. SAMPLED QUALITATIVE RESULTS

Figures (9)

  • Figure 1: Overview of LACE: Interface and Interaction
  • Figure 2: LACE’s modular design lets users configure or replace the underlying AI pipeline.
  • Figure 3: Layer-based prompting offers granular control over AI-generated elements.
  • Figure 4: Results from Task 1. Prompt: "A man reaching for a painting..."
  • Figure 5: Results from Task 2. Prompt: "An abstract composition that embodies..."
  • ...and 4 more figures