MAI Flash defaults to PCM: wrap it as WAV, vs Sume output formats

MAI-Voice-2.1-Flash on OpenRouter outputs mp3 or pcm and defaults to pcm, so a saved response will not play in most players until you add a WAV header. Sume's TTS Router defaults to mp3 and can return a ready-made wav.

The two defaults

The OpenRouter model page for Flash lists mp3 or pcm output with pcm as the default. The page I read does not state the sample rate in the text I used, so check it before you wrap the file. Sume documents the default as mp3 at 44.1 kHz and 128 kbps.

Wrap raw PCM as WAV

Python's standard wave module writes the header. Set the rate and sample width to what the service reports; the values here (24 kHz, 16-bit, mono) are placeholders, not Microsoft's documented numbers.

import wave

def pcm_to_wav(pcm_path, wav_path, rate=24000, width=2, channels=1):
    with open(pcm_path, "rb") as f:
        data = f.read()
    with wave.open(wav_path, "wb") as w:
        w.setnchannels(channels)
        w.setsampwidth(width)
        w.setframerate(rate)
        w.writeframes(data)

pcm_to_wav("flash.pcm", "flash.wav")

Ask Sume for wav directly

Set output_format to a container, sample rate and encoding and the job result is a normal wav. Wav files are large; a voiceover that must stay under an upload cap is smaller as mp3.

curl -sS -X POST https://api.sume.com/v1/tts-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"sonic-3.6","transcript":"Test line.","avatar_handle":"product_host","language":"en","output_format":{"container":"wav","sample_rate":24000,"encoding":"pcm_s16le"}}' \
  | jq -r '.data.request_id'

Where this stops

I did not call Flash, so the sample rate in the example is an assumption you must replace. If a header is wrong the audio plays fast, slow or as noise.