Codex GPT Image is the right call when Codex understands the project but there is no source still worth preserving. GPT Image 2 is the net-new and type-sensitive model in that loop, not a replacement for every visual request.

  • Codex already has the brief from README, tickets, or design notes.
  • GPT Image 2 is selected because the still must be invented or carry labels.
  • The file is written into the project, not left as a chat thumbnail.
  • Identity-preserving edits are routed elsewhere.

codex gpt image workflow overview

Media.io fit: Media.io CLI exposes GPT Image 2 inside Codex. Codex owns context and paths; Media.io returns the still.

In this article

What GPT Image 2 Is For Inside Codex

Use GPT Image in Codex when the job is a new hero, a labeled mock, or a poster-like frame. That is a different job from keeping a bottle identical.

codex gpt image - What GPT Image 2 Is For Inside Codex

If a source PNG must survive, do not default to GPT Image 2. Use Nano Banana 2 for that edit.

  • State size, use, and forbidden elements in the prompt file.
  • Do not pass a source image unless you mean to.
  • Write the result to a review filename.
  • Keep model choice visible in the job note.

Set Up Codex Before You Test GPT Image 2

A gpt image 2 Codex request cannot succeed if CLI, plugin, skills, or browser auth are missing. Probe with one still before you design a pack.

codex gpt image - Set Up Codex Before You Test GPT Image 2

Use the complete Codex setup prompt. Shortening INSTALL-HELP is how teams get a model name and no file.

Run the First Still Against Real Project Language

The first gpt image Codex job should use words from the repo: product name, audience, and where the still will live.

Codex already has Call Avoid
No source, need a new still GPT Image 2 or text to image Pretending an edit model will invent identity
A source that must stay Nano Banana 2 Dropping the source path

When you use GPT Image in Codex this way, the output is easier to accept because it already matches project vocabulary.

Save the File Where the Project Already Looks

Codex should write packaging-hero-recycled-box.jpg into the design or CMS folder, not a random download path.

codex gpt image - Run the First Still Against Real Project Language

  1. Pull constraints from the repo.
  2. Write a prompt file.
  3. Generate with GPT Image 2.
  4. Save beside the other campaign stills.

Reviewers open folders, not chat transcripts.

Use Media.io CLI so Codex Can Call GPT Image 2

After Media.io CLI is connected, Codex can send the brief to GPT Image 2 and receive a file path. Keep Codex image generation as the broader setup guide and this page as the model-specific route.

If You Want Codex to Set Media.io Up for You

Copy This Complete Setup Prompt into Codex

Copy-ready Codex setup prompt

Set up Media.io for me so I can generate images and videos from here.

  1. Install the CLI: run npm i -g @mediaio/cli.
  2. Install the companion plugin: run codex plugin marketplace add media-io/plugin, then run codex plugin add media-io@media-io.
  3. Install the companion skills: run npx skills add media-io/plugin -g.
  4. Authenticate: run mediaio auth login and complete the sign-in in the browser it opens.

If any install/update/sign-in step fails, read https://raw.githubusercontent.com/media-io/cli/refs/heads/main/INSTALL-HELP.md and follow it to diagnose and fix the problem automatically; only ask me for the steps you can't do yourself (browser sign-in, granting permissions).

Once that's done, let me know when it's ready.

Media.io CLI workflow for codex gpt image

Show a real terminal or agent session, the returned output path or status, and the generated result.

If the Still Is Weak, Fix the Brief Before the Model

Blurry type or a wrong product is usually a missing constraint, not a reason to abandon GPT Image 2.

codex gpt image - Save the File Where the Project Already Looks

Symptom First check
No file CLI, plugin, auth
Wrong product Brief missing locks
Looks like an ignored source You needed Nano Banana 2

Add the missing lock, then rerun the same job row.

FAQs About Codex Gpt Image

  • Can I use GPT Image in Codex natively?
    Codex can plan the still and choose a path, but pixels come from a connected model such as GPT Image 2 through Media.io CLI.
  • When should I use GPT Image 2 in Codex instead of Nano Banana 2?
    When there is no source still to preserve, or when labels and net-new composition matter more than identity.
  • What is GPT Image 2 Codex best at?
    Project-aware new stills, mock-style frames, and jobs where text rendering is part of the brief.
  • Where should the file go?
    Into the project folder reviewers already use, with a stable name that matches the shot.
  • Do I need a separate GPT Image website tab?
    No. After CLI, plugin, skills, and authentication, Codex can call the model in-session.
  • What if Codex returns a plan but no image?
    Check Media.io CLI setup, authentication, and that the job actually requested image generation.

Use GPT Image 2 When Codex Must Invent the Still

If the repository has context but no source file to protect, GPT Image 2 is the Codex image model to call. Keep identity edits on Nano Banana 2.

Nicola Massimo
Nicola Massimo Sep 24, 26
Share article:
media.io

AI Video Generator star

Easily generate videos from text or images

Generate