Use images as inputs or generated drafts and understand the different voice and video boundaries.
Conversation & research · Documented · Documentation checked 2026-10-04
ChatGPT Chat and Work; local Codex and the OpenAI API are separate surfaces.
Tools, usage, and controls vary by plan, client, region, rollout, and workspace policy.
The worked example is illustrative and has not been tested in a signed-in product account.
What it does and how it connects
Use images as inputs or generated drafts and understand the different voice and video boundaries.
Multimodal means interacting with more than text: an image can be an input, and image generation can produce or revise a visual draft. Supported desktop Voice can coordinate work. Image generation, voice interaction, and video generation are distinct capabilities; this guide does not infer video access from image access.
Walkthrough
Use the product named in the access panel. The steps below combine documented controls with the editorial example.
- Choose a permitted image or describe a new visual without personal identifiers.
- State whether the job is explanation, generation, or editing.
- Ask for a visual draft with a clear audience and intended use; inspect it before requesting revisions.
- Use available voice controls for a spoken discussion only after checking device and account support.
Worked example
Draft an illustration for a workshop showing a person reviewing an AI-generated recommendation. Use it as an editorial visual, then write a factual caption separately. Do not ask an image model to serve as the authoritative source of a precise technical diagram.
Create an editorial illustration for a beginner AI workshop: a person compares an AI suggestion with source documents before making a decision. Use midnight navy and warm amber. Avoid real people, logos, invented interface text, and factual claims inside the image. Then provide a plain-language caption that labels it as an illustration.
Inspect and verify the result
Inspect hands, faces, legibility, and unwanted symbols. Check the caption independently. For an input image, confirm that the explanation refers to visible details instead of imagined information outside the frame. Review a written summary of important spoken instructions.
Limits and troubleshooting
Image outputs can contain artifacts or misleading realism. Voice availability is surface-dependent and speech can be misheard. If an exact label or measurement matters, add it in a controlled editing tool after checking it.
Data, permissions, and actions
Images and audio can reveal faces, surroundings, screens, and other people. Obtain permission before sharing their material. A spoken request can still request consequential work, so keep publication, sending, and purchases explicitly reviewable.
Related reading
Skills, plugins, apps, and MCP
Separate reusable instructions from external tools and the authority granted to connected accounts.
From evidence to a beginner lesson
Create a sourced workshop lesson and an editable handout from approved material.
A weekly evidence digest with review
Repeat a public-source investigation and deliver a private draft digest.
Evaluations: does the system do the job?
Create representative tests for quality, tool behavior, refusal, escalation, and the effects of a change.
Sources and verification
Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.
Image generation · OpenAI
Using ChatGPT: Chat and Work · OpenAI