yum-archive/TaSTT-Whisper

High-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) model

git clone https://git.yummers.dev/yum-archive/TaSTT-Whisper

KonstantinSource codes8c4603c

master
branchesmasterall refs →
1.1 KiB37 linesraw
1namespace Whisper
2{
3	/// <summary>Flags for the audio capture</summary>
4	[Flags]
5	public enum eCaptureFlags: uint
6	{
7		/// <summary>No special flags</summary>
8		None = 0,
9		/// <summary>When the capture device supports stereo, keep stereo PCM samples in addition to mono</summary>
10		Stereo = 1,
11	}
12
13	/// <summary>Parameters for audio capture</summary>
14	public struct sCaptureParams
15	{
16		/// <summary>Minimum transcribe duration in seconds</summary>
17		public float minDuration;
18		/// <summary>Maximum transcribe duration in seconds</summary>
19		public float maxDuration;
20		/// <summary></summary>
21		public float dropStartSilence;
22		/// <summary></summary>
23		public float pauseDuration;
24		/// <summary>Flags for the audio capture</summary>
25		public eCaptureFlags flags;
26
27		/// <summary>Initialize the structure with some reasonable default values</summary>
28		public sCaptureParams()
29		{
30			minDuration = 7.0f;			// 7 seconds
31			maxDuration = 11.0f;		// 11 seconds
32			dropStartSilence = 0.25f;	// 250 ms
33			pauseDuration = 0.333f;		// 333 ms
34			flags = eCaptureFlags.None;
35		}
36	}
37}