Document MiMo v2.5 video/audio input handling and video sampling controls

[em2275]

Could you clarify the supported audiovisual input path for xiaomi-mimo-v2-5 on POST /api/v1/chat/completions?

The published chat schema documents an MP4 data URI in video_url.url and raw base64 WAV in input_audio.data with format: "wav". Xiaomi's direct MiMo documentation additionally describes fps and media_resolution on video content parts. These controls are not documented in Venice's chat video schema.

1. Does Venice pass MP4 video and a separate WAV together to MiMo's native visual and audio inputs? Is the MP4's own soundtrack also processed, or should clients supply the separate WAV?
2. How are source duration, frame timestamps and audio start offsets represented to the model? Are frames resampled or arranged into composite images?
3. Are MiMo's fps and media_resolution accepted and forwarded by Venice? If so, please document their exact placement, defaults, limits and effect on token accounting.
4. What WAV sample rate/channel layout is recommended, and is there response metadata that confirms which modalities were processed?

A minimal documented request example would help clients diagnose missed speech or incorrect event order without guessing unsupported parameters. This is a documentation question; no account credentials, user media or private conversation content are included.

References:

• Venice chat schema
• MiMo direct API video input guide