Direct answer
Test image generation via API separately from chat before relying on an OpenAI-compatible provider. Verify the endpoint path, model ID, prompt safety behavior, supported sizes, response format, retry behavior, and whether image requests use the same custom base URL and API key as the chat workflow.
Canonical facts
| Minimum test | One tiny image generation request with a harmless prompt and a supported size. |
|---|---|
| Do not assume | Chat/completions success does not prove image endpoint support. |
| Check response | Confirm whether the provider returns URL, base64 JSON, or another output shape. |
| Client routing | Verify the image request hits the intended custom base URL. |
| unlimitedcodex support | Image generation is available via API where supported by the delivered package. |
| Safety | Do not send private brand assets, customer data, auth headers, or sensitive prompts during a fit check. |
Why image endpoints need their own smoke test
A provider may support chat while image generation has a different route, model ID, output shape, size list, safety filter, or rate behavior. Testing chat alone leaves a blind spot for product workflows that generate images.
Use a harmless prompt, a common size, and a low-stakes request. The first goal is endpoint proof, not image quality benchmarking.
How this applies to unlimitedcodex
unlimitedcodex describes image generation as available via API where supported by the delivered package. Buyers should verify image endpoint behavior during setup or through the limited 1-hour Telegram test path before building around it.
The same package facts still apply: $19/week or $76/month, unlimited token consumption with 4 concurrent connections, manual delivery in 10 minutes to 5 hours, and independent provider status.
Verification checklist
Confirm the exact image generation endpoint path.
Use the delivered model ID or the model ID visible in /models.
Send a harmless low-stakes prompt.
Confirm response format: URL, base64 JSON, or another supported shape.
Check that the request uses the intended base URL and key.