Upload 5 files
Browse files- AudioResample.cs +21 -17
- README.md +1 -3
- audio_resample_22050_16000.sentis +0 -0
- audio_resample_44100_16000.sentis +0 -0
- info.json +2 -2
AudioResample.cs
CHANGED
@@ -21,17 +21,13 @@ public class AudioResample : MonoBehaviour
|
|
21 |
public bool playFinalAudio = true;
|
22 |
|
23 |
IWorker engine;
|
|
|
24 |
|
25 |
-
BackendType
|
26 |
-
|
27 |
-
Ops ops;
|
28 |
-
ITensorAllocator allocator;
|
29 |
|
30 |
void Start()
|
31 |
{
|
32 |
-
|
33 |
-
ops = WorkerFactory.CreateOps(backend, allocator);
|
34 |
-
|
35 |
ConvertAudio();
|
36 |
}
|
37 |
|
@@ -62,7 +58,7 @@ public class AudioResample : MonoBehaviour
|
|
62 |
return;
|
63 |
}
|
64 |
|
65 |
-
engine = WorkerFactory.CreateWorker(
|
66 |
|
67 |
int channels = inputAudio.channels;
|
68 |
int size = inputAudio.samples * channels;
|
@@ -76,15 +72,23 @@ public class AudioResample : MonoBehaviour
|
|
76 |
|
77 |
var output = engine.PeekOutput() as TensorFloat;
|
78 |
if (inputAudio.frequency == 44100)
|
79 |
-
{
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
}
|
85 |
else
|
86 |
{
|
87 |
-
output.
|
88 |
outData = output.ToReadOnlyArray();
|
89 |
}
|
90 |
|
@@ -103,7 +107,7 @@ public class AudioResample : MonoBehaviour
|
|
103 |
|
104 |
private void OnDestroy()
|
105 |
{
|
106 |
-
|
107 |
-
|
108 |
}
|
109 |
-
}
|
|
|
21 |
public bool playFinalAudio = true;
|
22 |
|
23 |
IWorker engine;
|
24 |
+
IBackend backend;
|
25 |
|
26 |
+
BackendType backendType = BackendType.GPUCompute;
|
|
|
|
|
|
|
27 |
|
28 |
void Start()
|
29 |
{
|
30 |
+
backend = WorkerFactory.CreateBackend(backendType);
|
|
|
|
|
31 |
ConvertAudio();
|
32 |
}
|
33 |
|
|
|
58 |
return;
|
59 |
}
|
60 |
|
61 |
+
engine = WorkerFactory.CreateWorker(backendType, model);
|
62 |
|
63 |
int channels = inputAudio.channels;
|
64 |
int size = inputAudio.samples * channels;
|
|
|
72 |
|
73 |
var output = engine.PeekOutput() as TensorFloat;
|
74 |
if (inputAudio.frequency == 44100)
|
75 |
+
{
|
76 |
+
//The model gives 2x as many samples as we would like so we fix it:
|
77 |
+
//We need to pad it if it has odd number of samples
|
78 |
+
int n = output.shape[1] % 2;
|
79 |
+
using var output2 = TensorFloat.AllocNoData(new TensorShape(1, output.shape[1] + n));
|
80 |
+
backend.Pad(output, output2, new int[] { 0, n }, Unity.Sentis.Layers.PadMode.Constant, 0);
|
81 |
+
|
82 |
+
//Now we take every second sample:
|
83 |
+
output2.Reshape(new TensorShape(output2.shape[1] / 2, 2));
|
84 |
+
using var output3 = TensorFloat.AllocNoData(new TensorShape(output2.shape[0], 1));
|
85 |
+
backend.Slice(output2, output3, new[] { 0 }, new[] { 1 }, new[] { 1 });
|
86 |
+
output3.CompleteOperationsAndDownload();
|
87 |
+
outData = output3.ToReadOnlyArray();
|
88 |
}
|
89 |
else
|
90 |
{
|
91 |
+
output.CompleteOperationsAndDownload();
|
92 |
outData = output.ToReadOnlyArray();
|
93 |
}
|
94 |
|
|
|
107 |
|
108 |
private void OnDestroy()
|
109 |
{
|
110 |
+
engine?.Dispose();
|
111 |
+
backend?.Dispose();
|
112 |
}
|
113 |
+
}
|
README.md
CHANGED
@@ -4,9 +4,7 @@ library_name: unity-sentis
|
|
4 |
pipeline_tag: audio-to-audio
|
5 |
---
|
6 |
|
7 |
-
# Model to turn 44kHz and 22kHz audio to 16kHz
|
8 |
-
*Version 1.3.0 Sentis files are not compatible with Sentis version 1.4.0 and need to be recreated/downloaded
|
9 |
-
|
10 |
This contains models in Sentis format to convert 44kHz and 22kHz audioclips to 16kHz. Useful for models such as [Whisper-Tiny](https://huggingface.co/unity/sentis-whisper-tiny)
|
11 |
|
12 |
## How to Use
|
|
|
4 |
pipeline_tag: audio-to-audio
|
5 |
---
|
6 |
|
7 |
+
# Model to turn 44kHz and 22kHz audio to 16kHz (Sentis version 1.4.0)
|
|
|
|
|
8 |
This contains models in Sentis format to convert 44kHz and 22kHz audioclips to 16kHz. Useful for models such as [Whisper-Tiny](https://huggingface.co/unity/sentis-whisper-tiny)
|
9 |
|
10 |
## How to Use
|
audio_resample_22050_16000.sentis
CHANGED
Binary files a/audio_resample_22050_16000.sentis and b/audio_resample_22050_16000.sentis differ
|
|
audio_resample_44100_16000.sentis
CHANGED
Binary files a/audio_resample_44100_16000.sentis and b/audio_resample_44100_16000.sentis differ
|
|
info.json
CHANGED
@@ -6,7 +6,7 @@
|
|
6 |
"audio_resample_22050_16000.sentis",
|
7 |
"audio_resample_44100_16000.sentis"
|
8 |
],
|
9 |
-
"version":
|
10 |
-
"1.
|
11 |
]
|
12 |
}
|
|
|
6 |
"audio_resample_22050_16000.sentis",
|
7 |
"audio_resample_44100_16000.sentis"
|
8 |
],
|
9 |
+
"version":[
|
10 |
+
"1.4.0"
|
11 |
]
|
12 |
}
|