[Paper Review] 3DALL-E: Integrating Text-to-Image AI in 3D Design Workflows

Vivian Liu, Jo Vermeulen+2 more|arXiv (Cornell University)|Oct 20, 2022

Design Education and Practice4 citations

TL;DR

3DALL-E integrates DALL-E, GPT-3, and CLIP into Fusion 360 to generate 2D image inspiration for 3D design workflows. Designers used text and image prompts to explore concepts, prevent fixation, and generate references, with 13 participants reporting strong potential for ideation and co-creation in CAD design.

ABSTRACT

Text-to-image AI are capable of generating novel images for inspiration, but their applications for 3D design workflows and how designers can build 3D models using AI-provided inspiration have not yet been explored. To investigate this, we integrated DALL-E, GPT-3, and CLIP within a CAD software in 3DALL-E, a plugin that generates 2D image inspiration for 3D design. 3DALL-E allows users to construct text and image prompts based on what they are modeling. In a study with 13 designers, we found that designers saw great potential in 3DALL-E within their workflows and could use text-to-image AI to produce reference images, prevent design fixation, and inspire design considerations. We elaborate on prompting patterns observed across 3D modeling tasks and provide measures of prompt complexity observed across participants. From our findings, we discuss how 3DALL-E can merge with existing generative design workflows and propose prompt bibliographies as a form of human-AI design history.

Motivation & Objective

• To explore how text-to-image AI can support conceptual 3D design workflows in CAD software.
• To investigate how designers use AI-generated image inspiration to overcome design fixation and spark new ideas.
• To evaluate the integration of multimodal AI (text and image prompts) within an industrial CAD environment like Fusion 360.
• To understand prompting patterns and complexity in real-world 3D design tasks.
• To propose prompt bibliographies as a form of human-AI design history for future collaboration.

Proposed method

• Integrated DALL-E, GPT-3, and CLIP into Fusion 360 via a plugin named 3DALL-E.
• Used GPT-3 to generate 3D keywords, design styles, and part suggestions based on user goals.
• Enabled image prompting by using a real-time render of the user’s 3D workspace as a visual input to DALL-E.
• Leveraged CLIP to align AI-generated suggestions with the current 3D model context, improving prompt relevance.
• Provided interactive, dynamic prompt suggestions that updated in real time as the user worked.
• Collected and analyzed user prompts and interactions during a 13-participant user study in real CAD workflows.

Figure 1 . 3DALL-E integrates a state-of-the-art text-to-image AI (DALL-E) into 3D CAD software Fusion 360. This plugin generates 2D image inspiration for conceptual CAD and product design workflows. 3DALL-E helps users craft text prompts by providing 3D keywords, design/styles, and parts from GPT-3

Experimental results

Research questions

• RQ1 How can text-to-image AI be effectively integrated into 3D CAD design workflows to support ideation?
• RQ2 What prompting patterns emerge when designers use AI-generated image inspiration during 3D modeling?
• RQ3 How does using image prompts derived from the current 3D model affect the quality and relevance of AI-generated outputs?
• RQ4 In what ways does AI-generated inspiration prevent design fixation and expand design considerations?
• RQ5 How can prompt histories be used to track and preserve human-AI collaborative design processes?

Key findings

• Designers found 3DALL-E valuable for generating reference images, preventing design fixation, and inspiring new design directions.
• Participants used diverse prompting strategies, including early, late, or iterative use of AI, depending on workflow stage.
• Prompt complexity varied significantly across users, with measurable differences in length and specificity.
• The system successfully bridged 3D modeling and text-to-image generation by using real-time viewport renders as image prompts.
• Designers reported that AI suggestions helped them access design language and 3D keywords they might not have otherwise considered.
• The authors propose prompt bibliographies as a novel method to preserve and reflect human-AI design history.

Figure 2 . 3DALL-E walkthrough. Step I: Initial state, where users can type their design intentions. Step II: Users are presented with prompt suggestions from GPT-3. Step III: Selected suggestions are rephrased into an editable prompt. Step IV: Users wait as DALL-E generates. Step V: Results are sho