#H

1 messages · Page 1 of 1 (latest)

zenith turret
#

Hi

unreal tartanBOT
#

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

#

Hello! I’m designed to answer questions specifically about OpenAI products, APIs, models, and related topics.

If you have a question about using the OpenAI API (e.g., GPT-4.1, GPT-4o, Assistants, Realtime, Whisper, pricing, rate limits, etc.), just ask and I’ll help.

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

#

@zenith turret Kapa.ai is still learning and improving, please let me know how I did by reacting below.

zenith turret
#

Hot take: gpt image 2 is garbage in the chatgpt app

#

@unreal tartan Hot take: gpt image 2 is garbage in the chatgpt app and image 1.5 is absolute garbage anywhere

unreal tartanBOT
#

Let me check the knowledge base for you again. One moment! :books:

#

Your experience is valid, and quality can definitely feel inconsistent, especially across different surfaces (ChatGPT vs API) and models.

From the official docs and guides:

  • OpenAI currently positions gpt-image-2 as the state-of-the-art and the recommended default for new builds, especially for high‑quality generation, editing, text-in-image, and brand‑sensitive work.[Model summary; GPT Image 2]
  • gpt-image-1.5 is described as the previous image model, mainly kept for existing workflows; for new work, OpenAI explicitly recommends moving to gpt-image-2.[Model summary; GPT Image 1.5]

There are acknowledged limitations that might match what you’re seeing:

  • Text rendering and layout can still be imperfect, even in the latest models.[Limitations]
  • Consistency and composition control (e.g., exact placement, recurring characters/brands) can be unreliable across generations.[Limitations]
  • Larger resolutions (above ~2K, e.g. > 2560×1440) are considered experimental, so quality can vary more.[Customize output]

If you’re open to it and want more predictable results, the docs suggest:

#
  • Use gpt-image-2 with quality: "high" and a “safe” size like 1024x1024, 1536x1024, or 1024x1536.[Model summary; Customize output]
  • Treat gpt-image-1.5 mainly as a legacy option while migrating prompts to gpt-image-2.[Upgrade path]

If you describe a specific failure case (e.g., “logo consistency”, “packaging text”, “photo edits”), I can point to the most relevant documented knobs (model, size, quality, parameters) that are actually supported and might help.

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

#

@zenith turret Kapa.ai is still learning and improving, please let me know how I did by reacting below.

zenith turret
#

@unreal tartan gpt image 2 is not able to do high in chatgpt app

unreal tartanBOT
#

Let me check the knowledge base for you again. One moment! :books:

#

The knowledge sources don’t explicitly describe which quality options are exposed inside the ChatGPT app UI, only what the underlying Image API and tools support.

From the docs we do know:

  • gpt-image-2 supports quality: "low" | "medium" | "high" | "auto" at the API level.[Customize output]
  • When used as a tool from models like gpt-5.5, the image generation tool also supports those quality options, but the app may choose defaults or hide some controls.[Image tool usage]

What’s not documented is:

  • Whether the ChatGPT app currently lets you explicitly choose "high" for gpt-image-2, or if it always uses "auto" or another default.
  • Any UI limitation specific to quality selection in the app.

So your observation that you “can’t do high in the ChatGPT app” is plausible, but I can’t confirm from these docs whether that’s by design, a UI limitation, or a bug. The only guaranteed way to force quality: "high" today is via the API (Images endpoint or image_generation tool), where that parameter is documented and supported.

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