Codex Desktop image editing produced no image output across repeated reference-image edit attempts
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 38/100
Research direction
The issue names no source files, tests, or entry points. Start by reproducing a reference-image edit in Codex Desktop on macOS and inspect the task history for the failed image-generation action. Done means the edit returns a preview or saved image, or exposes a structured actionable failure with the requested diagnostic details.
Written by the indexing model from the issue text.
Description
Summary
In Codex Desktop on macOS, repeated built-in image-edit requests using a reference cover image failed without producing any image output. From the user's perspective, image editing was unavailable: no preview, file path, or usable fallback image was ever returned.
This is related to #33050, but the reproduction here is specifically the reference-image editing path rather than text-to-image generation.
Environment
- Product: Codex Desktop
- Platform: macOS (Apple Silicon)
- Workflow: built-in image generation/editing in a local project thread
- Operation: edit an existing Chinese WeChat cover image
Requested edit
The task was intentionally narrow:
- preserve the existing illustrated person, dark background, torn grid-paper panel, and blue brush accents;
- replace only the title text with a new Chinese three-line title;
- produce one reviewable cover image.
Observed behavior
Two separate reference-image edit attempts failed:
- Create a title-free reusable cover master by removing existing title ink while preserving the rest of the image.
- Replace the existing title on the reference cover with a new Chinese three-line title.
Neither attempt produced a generated image, preview, saved path, or usable output artifact. The task history recorded the image-generation action as failed.
The agent then had to fall back to deterministic local text overlay, which was visually inferior for the requested poster-style Chinese typography and does not replace a functioning image-edit capability.
User impact
A user cannot treat the feature as available if repeated edits produce no image at all. Saying that image editing is theoretically supported is not useful when the visible outcome is zero generated images.
This also made the surrounding task much worse: the failed image action remained in the thread history, subsequent “continue” requests spent time recovering that failed path, and the thread later became unreliable.
Expected behavior
For a supported reference-image edit, Codex should either:
- return a generated image with a clear saved/preview location; or
- return a structured, actionable failure that distinguishes rate limiting, transport failure, service timeout, unsupported edit semantics, and moderation.
It should not leave the user waiting for a long time and then provide no image and no actionable diagnostic.
Requested improvements
- Surface the raw failure category and safe request correlation ID.
- Show an explicit timeout/retry policy for image edits.
- Clearly distinguish “image generation is running” from a stalled or failed request in the Desktop UI.
- Ensure a failed image request cannot silently poison later thread recovery.
- Consider a robust, deterministic text-replacement path for cover-image workflows where exact text and layout preservation matter.
No API keys, account identifiers, or source images are included in this report.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.5k
- Avg merge
- 1m
- Merged PRs (30d)
- 1k
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from openai/codex
-
enhancement remote
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
-
bug CLI windows-os
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
-
macOS sandbox blocks hw.optional.arm64 sysctl, causing Flutter to misdetect Apple Silicon as x64 Openbug CLI sandbox
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
-
bug CLI TUI
Difficulty 2/5 1-3 hours Newbie friendliness 90/100
-
CLI config enhancement skills
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 86/100
kwakseongjae/auto-hwp#319 ·
-
area:cli bug filter-quality good first issue priority:medium
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
-
Difficulty 1/5 Under an hour Newbie friendliness 72/100
bevyengine/bevy#25861 ·
-
comp-datalake
Difficulty 2/5 1-3 hours Newbie friendliness 88/100
ClickHouse/ClickHouse#121222 ·
-
A-linter
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
oxc-project/oxc#26863 ·