/v1/audio/voices returns plain strings instead of {id, name} objects #462

Bug

The /v1/audio/voices endpoint returns voices as a list of plain strings:

{"voices": ["af_alloy", "af_nova", ...]}

However, OpenAI-compatible clients (including Open WebUI) expect each voice to be an object with id and name fields:

{"voices": [{"id": "af_alloy", "name": "af_alloy"}, ...]}

This causes Open WebUI to silently fall back to its hardcoded list of 6 voices (alloy, echo, fable, nova, onyx, shimmer) instead of displaying all available Kokoro voices.

Steps to reproduce

1. Configure Open WebUI to use kokoro-fastapi as the OpenAI-compatible TTS backend
2. Open Admin > Settings > Audio and check the voice dropdown
3. Only 6 hardcoded voices appear instead of all available Kokoro voices

Expected behavior

All available Kokoro voices appear in the Open WebUI voice dropdown.

Fix

In api/src/routers/openai_compatible.py, change the list_voices return value from:

return {"voices": voices}

to:

return {"voices": [{"id": v, "name": v} for v in voices]}

Environment

• Image: ghcr.io/remsky/kokoro-fastapi-cpu:latest
• Open WebUI: ghcr.io/open-webui/open-webui:main