#gpt image

1 messages · Page 1 of 1 (latest)

stable robin
#

How can i make actual first person point of view of characters most of the time specifiying pov or fist person point of view from the eyes of "chrracter" wont work alone and delivers weird results

fervent krakenBOT
#

Hi @stable robin. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

#

To achieve a convincing first-person point of view (POV) with GPT Image models, simply specifying "POV" or "first person point of view from the eyes of [character]" is often not enough and can lead to inconsistent or strange results. The knowledge sources recommend a more descriptive and concrete approach to prompt engineering for visual perspective:

  • Describe the visual cues of first-person POV: Instead of just saying "first person," explicitly describe what would be visible from the character’s eyes. For example, mention the character’s hands, arms, or objects they are interacting with in the foreground, and describe the scene as it would appear from their vantage point.
  • Use detailed, concrete language: Specify the framing, angle, and what is in the foreground and background. For example:
    "You see your own hands holding a glowing sword, the blade pointed forward, with a dark forest stretching out ahead. The perspective is from your eyes, as if you are about to step into the woods."
  • Reference camera angles and framing: Borrow language from cinematography, such as "camera shot: from the character’s eye level, looking down at their hands," or "the viewer sees what the character sees, with their arms visible in the foreground" Visual cues that steer the look.
  • Iterate and refine: If the first result isn’t what you want, adjust your prompt to be even more explicit about what should be visible and how the scene should be framed. Iterative refinement is a best practice in prompt engineering Prompt engineering best practices.
#

Example strong prompt for first-person POV:

A fantasy scene from the first-person perspective: you see your own armored hands gripping the hilt of a sword, the blade pointed forward. In the distance, a dragon looms over a burning village. The viewpoint is from your eyes, as if you are about to charge.

By focusing on what the character would actually see and including their body parts or objects they interact with in the foreground, you guide the model to generate a more authentic first-person perspective. Avoid vague instructions and instead use concrete, visual details to steer the output Visual cues that steer the look.

If you continue to get unsatisfactory results, try rephrasing and adding more specific visual elements, as the model responds best to clear, descriptive prompts.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon: