Image editing regression + repeated 400 errors + inconsistent content filtering

Support Bug Reports
1

Summary:
Image editing in Cursor appears to have regressed significantly. The tool frequently regenerates new images instead of modifying the original, while also producing repeated “request failed with status code 400” errors. Additionally, content filtering behaviour is inconsistent and contradictory.

Expected Behaviour:

  • Image edits should preserve original composition, pose, and structure

  • The model should apply only the requested changes

  • Requests should complete without frequent 400 errors under normal usage

  • Content rules should be applied consistently to both prompts and outputs

Actual Behaviour:

  • The model often generates a completely new image instead of modifying the original

  • Composition and structure are not preserved

  • Repeated “request failed with status code 400” errors interrupt workflows

  • Content filtering appears inconsistent:

    • Prompts are sometimes rejected or altered due to restrictions (e.g. referencing firearms)

    • However, the model will still include similar elements in its own generated output

Steps to Reproduce:

  1. Upload an existing image

  2. Request a simple edit (e.g. clothing change, colour adjustment, or adding small elements)

  3. Observe that:

    • A new image is generated instead of an edit

    • OR a 400 error occurs

    • OR the output includes elements that were restricted in the prompt

Additional Notes:

  • This behaviour differs from previous usage where edits were more reliable and composition was preserved

  • Reattaching the image and simplifying prompts does not consistently resolve the issue

  • The issue affects both simple and more detailed edit requests

Impact:

  • Breaks iterative workflows (e.g. concept art refinement)

  • Causes loss of control over outputs

  • Requires repeated retries due to errors

  • Reduces trust in the tool due to inconsistent rule enforcement

Request:

  • Restore reliable image-to-image editing behaviour

  • Fix repeated 400 error occurrences during normal usage

  • Ensure consistent application of content rules between prompt handling and generated output

  • Improve adherence to explicit edit constraints (e.g. “preserve composition, modify only X”)

Happy to provide examples if needed.

2

Hi there!

We detected that this may be a bug report, so we’ve moved your post to the Bug Reports category.

To help us investigate and fix this faster, could you edit your original post to include the details from the template below?

Bug Report Template - Click to expand

Where does the bug appear (feature/product)?

  • Cursor IDE
  • Cursor CLI
  • Background Agent (GitHub, Slack, Web, Linear)
  • BugBot
  • Somewhere else…

Describe the Bug
A clear and concise description of what the bug is.

Steps to Reproduce
How can you reproduce this bug? We have a much better chance at fixing issues if we can reproduce them!

  1. …
  2. …
  3. …

Expected Behavior
What is meant to happen here that isn’t working correctly?

Screenshots / Screen Recordings
If applicable, attach images or videos (.jpg, .png, .gif, .mp4, .mov)

Operating System

  • Windows 10/11
  • MacOS
  • Linux

Version Information

  • For Cursor IDE: Menu → About Cursor → Copy
  • For Cursor CLI: Run agent about in your terminal
IDE:
Version: 2.xx.x
VSCode Version: 1.105.1
Commit: ......

CLI:
CLI Version 2026.01.17-d239e66

For AI issues: which model did you use?
Model name (e.g., Sonnet 4, Tab…)

For AI issues: add Request ID with privacy disabled
Request ID: f9a7046a-279b-47e5-ab48-6e8dc12daba1
For Background Agent issues, also post the ID: bc-…

Additional Information
Add any other context about the problem here.

Does this stop you from using Cursor?

  • Yes - Cursor is unusable
  • Sometimes - I can sometimes use Cursor
  • No - Cursor works, but with this issue

The more details you provide, the easier it is for us to reproduce and fix the issue. Thanks!

5

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Image editing/generation appears to ignore the source image composition and repeatedly fails with status code 400 before eventually generating an unrelated replacement image.

The system previously handled image-to-image edits much more accurately, preserving composition and only modifying requested areas. Recent behaviour appears significantly more restrictive or unstable.

Steps to Reproduce

  1. Upload a base image.

  2. Upload a secondary reference image.

  3. Ask Cursor to preserve the original art style and composition while adding creatures/elements from the second image into the environment.

  4. Generate the image.

Prompt used:

“Using the first image following the same art style guide can you add alien creatures referenced in the second image to the surrounding jungle environment in the first image hiding in cover with subtle hints to their presence? Maintain the first image created with the characters intact, only add additional creatures to the surrounding environment.”

Expected Behavior

The original image composition and characters should remain intact while only adding subtle environmental creatures/elements inspired by the second image.

The system should perform a controlled image edit rather than generating an entirely new composition.

Actual Behaviour

The generation repeatedly failed with:
“Request failed with status code 400”

After multiple failures, Cursor generated an entirely different image with:

  • different composition

  • different character placement

  • different art structure

  • poor preservation of the original uploaded image and art style

It appears the system fell back to full image regeneration instead of image editing.

Operating System

MacOS

Version Information

3.2.16

For AI issues: which model did you use?

Gemini 3.1 Pro

The same issue also occurs with other models.

Request ID

f9a7046a-279b-47e5-ab48-6e8dc12daba1

Additional Information

This behaviour appears significantly worse than previous Cursor image-generation behaviour from earlier versions/builds. Earlier versions preserved uploaded image structure much more reliably.

The issue seems particularly noticeable when:

  • using two image references

  • requesting subtle edits only

  • asking to preserve composition/layout

  • attempting environment-only modifications

The system often appears to abandon the original image edit request and instead generates an entirely new image loosely based on keywords from the prompt.

At present this issue is severe enough that it is pushing me toward using alternative AI image-generation systems instead of Cursor for this workflow.

Does this stop you from using Cursor?

Yes - Cursor is unusable

6

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Over the past several days, Cursor’s image editing behaviour appears to have changed significantly, particularly when using multiple image references and requesting localised edits while preserving composition.

Previously, the system handled:

composition preservation

environment-only modifications

subtle additions to existing artwork

reference-guided edits

…with relatively high fidelity.

Current behaviour now frequently results in:

repeated 400 request failures

ignored composition/layout

full scene regeneration instead of controlled edits

loss of original character positioning

unrelated image outputs loosely derived from prompt keywords

Observed Failure Pattern

The issue appears most reproducible when:

using two reference images

requesting preservation of the original composition

asking for environmental additions only

specifying “maintain characters intact”

performing subtle img2img edits rather than full generations

Example prompt:

“Using the first image following the same art style guide can you add alien creatures referenced in the second image to the surrounding jungle environment in the first image hiding in cover with subtle hints to their presence? Maintain the first image created with the characters intact, only add additional creatures to the surrounding environment.”

Suspected Behaviour

The pipeline appears to:

attempt image editing

fail repeatedly (400)

abandon img2img preservation

fall back to full image regeneration

The resulting image often:

changes composition entirely

relocates characters

alters scene structure

ignores the uploaded base image

This behaviour feels substantially different from previous Cursor builds where composition retention was far more reliable.

Additional Notes

This issue appears across multiple models, not only Gemini 3.1 Pro.

The degradation specifically affects workflows involving:

concept art

sketch refinement

composition-preserving edits

iterative environment detailing

At present the workflow reliability has dropped enough that alternative platforms are becoming more viable for image editing tasks.

Would appreciate clarification on whether:

image editing behaviour changed recently

moderation/routing systems were updated

img2img preservation was altered internally

reference-image handling is currently degraded

Steps to Reproduce

Open Cursor image generation/editing.
Upload a base artwork image containing an established composition and character placement.
Upload a second reference image containing creature/environment reference material.
Use a prompt requesting:
preservation of the original image composition
preservation of existing characters
subtle environmental additions only
matching of the original art style
Example prompt used:

“Using the first image following the same art style guide can you add alien creatures referenced in the second image to the surrounding jungle environment in the first image hiding in cover with subtle hints to their presence? Maintain the first image created with the characters intact, only add additional creatures to the surrounding environment.”

Generate the image.
Observe:
repeated 400 request failures
eventual fallback generation
composition/layout changes
poor preservation of uploaded source image
full image regeneration instead of localised editing

Expected Behavior

The system should perform a controlled image edit using the uploaded base image as the primary composition reference.

Expected output behaviour:

preserve the original image composition and camera framing
preserve existing character placement and anatomy
maintain the original art style
apply only the requested environmental modifications
add subtle creature presence to surrounding jungle areas without restructuring the scene

The generation should behave as an image-editing/img2img operation rather than a full scene regeneration.

The system should also complete the request without repeated 400 failures or fallback behaviour that ignores the uploaded source image.

Operating System

MacOS

Version Information

Version: 3.2.16 (Universal)
VSCode Version: 1.105.1
Commit: 3e548838cf824b70851dd3ef27d0c6aae371b3f0
Date: 2026-04-28T21:07:47.682Z (4 days ago)
Layout: glass
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin x64 24.6.0

For AI issues: which model did you use?

Gemini 3.1 Pro

Does this stop you from using Cursor

No - Cursor works, but with this issue

7

The main point of confusion on my side came from the fact that earlier outputs were significantly closer to the source image composition and structure, which gave the impression that a more direct image-editing workflow was in place. Even if that was technically emergent rather than designed behaviour, it materially affected how usable the tool felt for iterative concept work.

What’s currently limiting the workflow is less the “text-to-image vs img2img” distinction itself, and more the loss of strong structural adherence to the uploaded reference image when trying to do controlled, incremental scene changes.

From a practical standpoint, that shift has a big impact on:

  • composition preservation workflows

  • incremental environment building

  • sketch-to-final refinement

  • multi-reference scene augmentation

Just to clarify expectations going forward, is there any intention to:

  • improve reference-image structural weighting in future updates

  • introduce true inpainting / region-based editing capabilities

  • or stabilise the current behaviour so composition adherence is more consistent, even within the existing generation model?

Even a rough indication of whether this is likely to improve or is considered out of scope would be helpful for deciding whether to adapt workflows around Cursor or move that part of the pipeline elsewhere.

Appreciate the explanation so far non the less.

8

Thanks for the clarification on the underlying image generation approach.

What I’m trying to understand is whether there have been any recent changes (intentional or otherwise) to:

  • prompt interpretation or weighting

  • reference image influence

  • safety/filtering behaviour affecting image generation

  • or backend model routing/settings

From a user perspective, the system has changed quite noticeably in the last few days/weeks compared to how it performed over the past few months. Even with the same general prompting approach, outputs now appear less consistent, less aligned with reference structure, and more prone to full regeneration rather than guided variation.

If there have been updates in any of these areas, could you confirm whether:

  • this is an expected change in behaviour going forward, or

  • whether there are planned improvements/fixes to restore stronger prompt adherence and reference consistency

I want to be direct here: the previous behaviour was reliable enough for my workflow around iterative concept art and composition-based generation. At the moment, the current behaviour makes that workflow effectively unusable in practice, even if the underlying system is working as intended.

I’m mainly trying to understand whether I should adapt to this new baseline or expect further refinement that would restore that level of control.