Feature request: Bridge ChatGPT visual/image capabilities with Codex for game-asset workflows

Problem

For game-development workflows, ChatGPT is often much better at understanding visual references and generating images, while Codex is much better at editing the actual project and writing code. Today these feel too disconnected.

A concrete example is a Unity game project that needs a consistent character/sprite pipeline. The user can show ChatGPT a visual target and discuss art direction, sprite layering, 8-direction animation, SPR/ACT-style data, etc. ChatGPT can reason about the image and generate/iterate on visuals, but Codex working inside the project does not automatically inherit that visual understanding or the generated design decisions.

This creates a frustrating gap:

• ChatGPT can analyze the reference image well, but cannot directly continue the implementation inside the Codex project workspace.
• Codex can modify the Unity project, but its visual/image generation and visual analysis workflow is weaker or disconnected from what ChatGPT already understood.
• The user has to manually copy prompts, screenshots, conclusions, and requirements between ChatGPT and Codex.
• Context and design intent can be lost between the two, which leads to rework.

Requested improvement

Please make Codex either stronger at image generation/visual analysis, or provide a first-class bridge between ChatGPT and Codex.

Ideal behavior:

1. ChatGPT and Codex can share the same task/project context when the user explicitly links them.
2. Codex can ask ChatGPT's image/vision system to analyze a reference image or generate an asset, then receive the result back in a structured way.
3. ChatGPT can hand an implementation task to Codex with the full design context, image references, constraints, and accepted decisions preserved.
4. Codex can return code changes, screenshots/renders, test results, and project state back to ChatGPT for review.
5. Both sides should understand which local project/folder/repository the task belongs to so users with multiple projects/accounts do not accidentally modify the wrong workspace.
6. Generated assets should be attachable to the active project workflow without manual downloading/re-uploading between products.

Why this matters

For game development, image/design and implementation are tightly coupled. A typical loop is:

visual reference -> art/design discussion -> generated concept -> implementation -> in-engine preview -> visual review -> code/art revision

ChatGPT is already strong at the visual/design side, while Codex is strong at project/code execution. A reliable bridge between them would be substantially more useful than treating them as isolated assistants.

Suggested workflow

A user should be able to say in ChatGPT:

Use this visual target for the MYTHWEAVE character system. Hand the implementation to Codex in my Unity project, then bring the result back here for review.

Then Codex could work on the project, generate an in-engine preview or screenshot, and return that result to ChatGPT, where the user can continue the visual review without manually rebuilding the context.

This would be especially valuable for Unity/Unreal/Blender/game-asset pipelines, UI implementation, shaders, sprite systems, and other workflows where code and visual output must be iterated together.

No files, tests, or implementation entry points are identified. Start by reviewing the Codex project architecture and the requested ChatGPT handoff workflow, including shared context, visual inputs and outputs, project selection, and asset transfer. Done would require a defined, reliable bridge that preserves design context and returns implementation results for review.