Image and Audio APIs

CoreRouter supports OpenAI-compatible image and audio interfaces. Actual capabilities, sizes, voices, file sizes, and format limits depend on models and channels; refer to console descriptions.

API Overview

Capability	Endpoint	Request Format	Key Fields
Image Generation	/v1/images/generations	JSON	model, prompt, size, n
Image Editing	/v1/images/edits	multipart/form-data or JSON	model, image, prompt
Legacy Image Edit Compatibility Path	/v1/edits	JSON	Legacy client compatibility; prefer /v1/images/edits when uploading files
Audio Transcription	/v1/audio/transcriptions	multipart/form-data	model, file
Audio Translation	/v1/audio/translations	multipart/form-data	model, file
Text-to-Speech	/v1/audio/speech	JSON	model, input, voice

Image Generation

export COREROUTER_API_KEY="sk-xxxxxxxxxxxxxxxx"

curl https://api.corerouter.cloud/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $COREROUTER_API_KEY" \
  -d '{
    "model": "image-model-id",
    "prompt": "A clean tech product documentation cover, dark background, blue-green light effects",
    "size": "1024x1024",
    "n": 1
  }'

Image Editing

Image editing typically uses multipart/form-data:

curl https://api.corerouter.cloud/v1/images/edits \
  -H "Authorization: Bearer $COREROUTER_API_KEY" \
  -F "model=image-model-id" \
  -F "[email protected]" \
  -F "prompt=Change background to dark tech style, keep subject"

/v1/images/edits can also use JSON per client protocol, but images must be passed as URLs or Base64 data supported by that protocol. Don't convert the same multipart request to /v1/edits; the legacy path won't process it as a new multipart image edit flow.

Audio Transcription

curl https://api.corerouter.cloud/v1/audio/transcriptions \
  -H "Authorization: Bearer $COREROUTER_API_KEY" \
  -F "model=audio-model-id" \
  -F "[email protected]"

Audio Translation

curl https://api.corerouter.cloud/v1/audio/translations \
  -H "Authorization: Bearer $COREROUTER_API_KEY" \
  -F "model=audio-model-id" \
  -F "[email protected]"

Text-to-Speech

curl https://api.corerouter.cloud/v1/audio/speech \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $COREROUTER_API_KEY" \
  -d '{
    "model": "tts-model-id",
    "voice": "alloy",
    "input": "Hello, welcome to CoreRouter."
  }' \
  --output speech.mp3

Usage Recommendations

• Image count, size, quality, and audio duration all affect costs; test on a small scale first.
• n in image generation indicates the number of images. Don't allow users to pass unlimited n in your application.
• Image sizes use the letter x, e.g., 1024x1024, not the multiplication sign ×.
• When providing upload interfaces, limit file size, format, and duration.
• Don't reuse chat model IDs for image and audio models unless the console explicitly indicates support.
• If SDK wrapping of media interfaces is incomplete, directly use curl or HTTP clients.
• TTS returns are typically audio bytes; curl needs --output to save files.
• Voice recognition and translation typically require file upload, not just passing file path strings to JSON.

Common Issues

• 400: Check if multipart/form-data field names are correct and file paths exist.
• 415: Check if file format is supported by the model.
• Empty file returned: Confirm command has --output and check if response Header is audio content.
• Abnormal costs: Check if n, size, audio duration, and quality parameters meet expectations.