Skip to content

Docs/UX: Realtime Talk with OpenAI requires OpenAI Platform API credits, not Codex/ChatGPT subscription quota #76498

@Hal2600

Description

@Hal2600

Summary

Realtime Talk with the OpenAI provider failed with an OpenAI insufficient_quota / You exceeded your current quota error even though Codex OAuth was working for normal OpenClaw chat/model turns and the ChatGPT/Codex usage dashboard still showed remaining quota.

After troubleshooting, the missing piece was that OpenAI Realtime Talk uses the OpenAI Platform API billing/key path, not ChatGPT/Codex subscription quota.

Environment

  • OpenClaw version: 2026.5.2
  • Platform: macOS
  • Feature: Talk / realtime voice
  • Talk provider: openai
  • Working Talk model after fix: gpt-realtime-mini
  • Normal chat/model route: Codex OAuth / openai-codex/* was present and working separately

What happened

  1. Talk was configured with talk.provider: "openai".
  2. Normal OpenClaw chat/model turns had Codex OAuth available.
  3. Talk failed with:
You exceeded your current quota, please check your plan and billing details.
  1. The user's ChatGPT/Codex usage buckets still showed remaining quota, which made the error confusing.
  2. Attempting to force Talk to openai-codex failed because that is not a registered realtime voice provider:
Realtime voice provider "openai-codex" is not registered
  1. Installed realtime Talk providers were OpenAI and Google, not openai-codex.
  2. Adding/funding OpenAI Platform billing and replacing the OPENAI_API_KEY used by OpenClaw fixed Talk. Realtime voice then worked with gpt-realtime-mini.

Why this was confusing

For users who sign into OpenClaw with Codex/ChatGPT OAuth, it is natural to assume that a new OpenAI-powered Talk feature might use the same subscription/usage bucket as normal OpenClaw chat turns.

The quota error sounds like the upgraded OpenAI/Codex account is out of quota, when the actual issue may be that the separate OpenAI Platform API key/project is unfunded, missing, or not the key OpenClaw is using.

Suggestion

Please consider adding a clearer note in Talk setup, docs, or the error handling path, such as:

OpenAI realtime Talk uses OpenAI Platform API billing and requires a funded OPENAI_API_KEY. ChatGPT Plus/Pro/Codex subscription quota and openai-codex/* OAuth do not fund realtime Talk audio.

It may also help if the Talk setup UI/error message distinguishes:

  • normal chat/model auth via Codex OAuth
  • OpenAI Platform API key billing for realtime audio
  • openai-codex not being a realtime voice provider

This would save new Talk users a lot of troubleshooting time, especially as the realtime feature is new and easy to confuse with Codex subscription usage.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions