Skip to content
TheChatArchiveThe Chat Archive
o4-mini-2025-04-16· OpenAIHugging Face LMArena Human Preference 140K

If I asked a ai chat model to generate an image of someone writing with their left hand,…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Mistralmistral-medium-2505’s answer.
  1. User

    If I asked a ai chat model to generate an image of someone writing with their left hand, could they do it? Would they generate an image that follows my commands? Would the person in the generated image actually use their left hand instead of the right?

  2. Assistant

    Yes—in principle, any modern text-to-image model (DALL·E, Stable Diffusion, Midjourney, etc.) can obey a prompt like “a person writing with their left hand.” Here’s how it typically plays out:

    1. Can they do it?
      • Most models have seen enough examples of people writing to generate a plausible scene.
      • If you explicitly say “left hand,” the network will try to put the pen or pencil in the model’s left hand and orient the arm/torso accordingly.

    2. Will they follow your command?
      • Generally, yes—especially if you’re clear and add emphasis (“a left-handed writer,” “the person is holding the pen in their left hand,” “left hand visible writing on paper”).
      • Occasionally you’ll get a slip (e.g. the model draws it in the right hand, or the anatomy is weird). Prompt-engineering (adding more detail) or trying a few variations usually fixes it.

    3. Is there a “real” left-hander behind the image?
      • No. It’s a purely synthetic scene. The “person” doesn’t exist outside the pixels. So there’s no actual human doing the writing—just an AI’s best guess at what left-handed writing looks like.

    Tips for better results
    • Be explicit: “a close-up of a left hand gripping a pen, writing ‘Hello’ on lined paper.”
    • Show context: mention desk angle, notebook, background lighting.
    • Iterate: if the first try still shows a right hand, tweak the phrasing or add synonyms (“southpaw,” “handedness,” “left-side writing”).

    Have you tried prompting one of these models yet? If so, what did your results look like? 😊

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