Images, audio, video and Realtime
Confirm model and channel capabilities before selecting an endpoint.
Availability
Actual capabilities depend on your account, model and channel configuration. First query /v1/models, then confirm the chosen model's endpoints, parameters, pricing and quotas through the console or support.
| Capability | Endpoint | Requirements |
|---|---|---|
| Image generation | POST /v1/images/generations | Compatible model, channel or configured task plugin |
| Image editing | POST /v1/images/edits | Confirm multipart/JSON format, reference images, masks, size and output format |
| Transcription | POST /v1/audio/transcriptions | Appropriate audio model; upload format and size vary by channel |
| Audio translation | POST /v1/audio/translations | Confirm support in the target model |
| Speech synthesis | POST /v1/audio/speech | Appropriate TTS model; voices, encoding and streaming vary by channel |
| Video tasks | POST /v1/videos, GET /v1/videos/:task_id | Compatible video channel/task plugin; providers do not necessarily share parameters |
| Video results | GET /v1/videos/:task_id/content | An authorized, completed task; availability depends on its task protocol |
| Realtime | wss://api.aimirrorx.com/v1/realtime?model=MODEL_ID | A Realtime-compatible channel, model and WebSocket client |
| Responses WebSocket | wss://api.aimirrorx.com/v1/responses | Uses the separate Responses protocol, not the Realtime audio protocol |
Image example
Confirm that your model supports the OpenAI Images protocol before running:
curl https://api.aimirrorx.com/v1/images/generations \
-H "Authorization: Bearer $AIMIRRORX_API_KEY" \
-H 'Content-Type: application/json' \
-d '{"model":"MODEL_ID_FROM_V1_MODELS","prompt":"一座安静的海边灯塔","n":1}'Results may contain a URL, base64 data or an asynchronous task object, depending on the protocol and channel. Successful submission does not mean generation is complete. Do not automatically resubmit a task with an uncertain outcome. Expiry, result downloads, polling frequency and billing follow the selected service's actual behavior.
Audio and video
Audio uploads typically use multipart, while speech synthesis typically returns binary data. Do not parse every response as text JSON. Video generation is a task flow: submit, save the task ID, poll its status, then download the result. Native provider paths can differ; use the protocol explicitly provided for your channel. This guide does not offer a universal video-parameter template or promise availability.
WebSocket and browsers
Server-side WebSocket clients can send Bearer authentication as required by the endpoint. A browser's native WebSocket API cannot set arbitrary request headers. Do not expose long-lived API keys in webpage code or URLs. Use a controlled server-side integration; do not assume support for provider-native temporary session keys, WebRTC or every event type.
Unsupported assumptions
Some upstream endpoints, including /v1/images/variations, Files and Fine-tunes, are currently unsupported. Administration APIs require separate administrative permissions; do not use customer tokens to attempt channel, user or billing administration. The upstream plugin marketplace is reference material. The documentation service does not install plugins or proxy API execution.