Re: [PATCH v5 0/12] MR1948: Draft: windows.media.speech: Recognition session work the second.

[Jinoh Kang (＠iamahuman)]

Jinoh Kang (@iamahuman) commented about dlls/windows.media.speech/recognizer.c:
> +
> +    if (FAILED(hr = IMMDevice_Activate(mm_device, &IID_IAudioClient, CLSCTX_INPROC_SERVER, NULL, (void**)&session->audio_client)))
> +        goto cleanup;
> +
> +    if (SUCCEEDED(hr = IMMDevice_GetId(mm_device, &str)))
> +    {
> +        TRACE("selected capture device ID: %s\n", debugstr_w(str));
> +        CoTaskMemFree(str);
> +    }
> +
> +    if (FAILED(hr = IAudioClient_GetMixFormat(session->audio_client, (WAVEFORMATEX **)&wfx)))
> +        goto cleanup;
> +
> +    wfx->wFormatTag = WAVE_FORMAT_PCM;
> +    wfx->nChannels = 1;
> +    wfx->nSamplesPerSec = 16000;
> Magic constant. You should replace this with a `#define` shared with Unix side interfacing vosk.  `#define WINE_VOSK_SAMPLE_RATE 16000` will do.

(I'm aware that most vosk models are trained with 16kHz PCM streams.)

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/1948#note_21044