Support Xiaomi MiMo `mimo-v2-tts` on OpenAI-compatible `/v1/audio/speech`

Read This First (Do Not Remove This Section)

• Docs: https://docs.newapi.ai/
• Usage questions first: https://deepwiki.com/QuantumNous/new-api
• Warning: issues with this template removed, section headings deleted, or content cleared may be closed directly. Repeated abusive submissions may result in a block.

Your current newapi version 0.11.5

Submission Checks

• I have confirmed there are no similar issues
• I have thoroughly read the docs at https://docs.newapi.ai/ and the project README, and confirmed the current version cannot meet my needs
• I have not removed any guidance or section headings from this template and will complete it as requested
• I understand that maintainers have limited time and issues that do not follow this template may be ignored or closed directly

Feature Description Add compatibility support for Xiaomi MiMo mimo-v2-tts in the OpenAI-compatible /v1/audio/speech endpoint by internally bridging to /v1/chat/completions.

Use Case Allow OpenAI-compatible clients (e.g., Open WebUI) that rely on /v1/audio/speech to use MiMo TTS models without modifying client-side code.

Background

Some clients, such as Open WebUI, are fixed to the OpenAI-compatible /v1/audio/speech API for TTS.

However, Xiaomi MiMo mimo-v2-tts only provides TTS through /v1/chat/completions with an audio field, so it cannot be used directly through the current OpenAI audio relay path.

Problem

Even if a channel uses https://api.xiaomimimo.com and the upstream model is mimo-v2-tts, the current /v1/audio/speech relay path still forwards to upstream /v1/audio/speech, which MiMo does not support.

As a result, OpenAI-compatible TTS clients cannot use MiMo TTS through new-api.

Proposed solution

Add a MiMo-specific compatibility bridge in the OpenAI audio relay path:

• detect upstream model mimo-v2-tts
• rewrite upstream request URL from /v1/audio/speech to /v1/chat/completions
• convert OpenAI audio speech requests into MiMo chat+audio payload
• use api-key authentication for MiMo
• decode the audio payload from the MiMo chat response and return a normal audio response to the client

Scope / compatibility

• only enabled when upstream model is mimo-v2-tts
• existing OpenAI TTS behavior remains unchanged
• existing non-MiMo channels are unaffected
• streaming /v1/audio/speech compatibility is not included

Notes

MiMo TTS appears to expect the target synthesis text in the assistant message content, while the user message can be used as an optional instruction/prompt for style guidance. The compatibility layer can adapt this automatically for OpenAI-style TTS callers.

If this direction is acceptable, I can open a PR with implementation and tests.