Skip to content
TheChatArchiveThe Chat Archive
Claudeclaude-3-5-haiku-20241022· AnthropicHugging Face LMArena Human Preference 140K

how would you improve this ideogram universal prompt to get the best output image from…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with gpt-4.1-2025-04-14’s answer.
  1. User

    how would you improve this ideogram universal prompt to get the best output image from ideogram using their Magic prompt feature??

    Cute vector illustration of a cute graphic that represents the funny slogan "insert slogan here". vector design but with thin dark outlines, vibrant colors, isolated on a black background.

  2. Claude Assistant

    I can help you refine the prompt to potentially get a better output from Ideogram's Magic Prompt feature. Here's an improved version:

    Whimsical vector illustration with a playful cartoon-style graphic embodying the humorous slogan "insert slogan here". Crisp, thin dark outlines defining clean shapes, using a vibrant color palette with high contrast. Stylized design with a minimalist approach, featuring bold and saturated colors that pop against a solid black background. Cute, modern graphic design aesthetic with a touch of humor and graphic simplicity. Ensure clean lines, sharp edges, and a contemporary vector art style that feels both fun and visually striking.
    

    Key improvements:

    • Added more descriptive language about the design style
    • Specified the cartoon/vector aesthetic more precisely
    • Emphasized the color contrast and graphic design elements
    • Provided more context about the visual mood and approach
    • Included specific art direction cues
    • Maintained the core request of the original prompt
    • Expanded on the visual characteristics

    The additional details should help Ideogram's AI better understand the specific visual style you're looking for, potentially resulting in a more targeted and refined image generation.

    Would you like me to elaborate on any part of the prompt or discuss vector illustration techniques further?

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.