Spaces:
Sleeping
Sleeping
Upload folder using huggingface_hub
Browse files- app.py +64 -0
- default-favicon.ico +0 -0
- requirements.txt +3 -0
app.py
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from fastrtc import (
|
| 2 |
+
Stream,
|
| 3 |
+
AdditionalOutputs,
|
| 4 |
+
audio_to_float32,
|
| 5 |
+
ReplyOnPause,
|
| 6 |
+
get_twilio_turn_credentials,
|
| 7 |
+
)
|
| 8 |
+
from functools import lru_cache
|
| 9 |
+
import gradio as gr
|
| 10 |
+
from typing import Generator, Literal
|
| 11 |
+
from numpy.typing import NDArray
|
| 12 |
+
import numpy as np
|
| 13 |
+
from moonshine_onnx import MoonshineOnnxModel, load_tokenizer
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
@lru_cache(maxsize=None)
|
| 17 |
+
def load_moonshine(
|
| 18 |
+
model_name: Literal["moonshine/base", "moonshine/tiny"],
|
| 19 |
+
) -> MoonshineOnnxModel:
|
| 20 |
+
return MoonshineOnnxModel(model_name=model_name)
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
tokenizer = load_tokenizer()
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
def stt(
|
| 27 |
+
audio: tuple[int, NDArray[np.int16 | np.float32]],
|
| 28 |
+
model_name: Literal["moonshine/base", "moonshine/tiny"],
|
| 29 |
+
) -> Generator[AdditionalOutputs, None, None]:
|
| 30 |
+
moonshine = load_moonshine(model_name)
|
| 31 |
+
sr, audio_np = audio # type: ignore
|
| 32 |
+
if audio_np.dtype == np.int16:
|
| 33 |
+
audio_np = audio_to_float32(audio)
|
| 34 |
+
if audio_np.ndim == 1:
|
| 35 |
+
audio_np = audio_np.reshape(1, -1)
|
| 36 |
+
tokens = moonshine.generate(audio_np)
|
| 37 |
+
yield AdditionalOutputs(tokenizer.decode_batch(tokens)[0])
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
stream = Stream(
|
| 41 |
+
ReplyOnPause(stt, input_sample_rate=16000),
|
| 42 |
+
modality="audio",
|
| 43 |
+
mode="send",
|
| 44 |
+
ui_args={
|
| 45 |
+
"title": "Live Captions by Moonshine",
|
| 46 |
+
"icon": "default-favicon.ico",
|
| 47 |
+
"icon_button_color": "#5c5c5c",
|
| 48 |
+
"pulse_color": "#a7c6fc",
|
| 49 |
+
"icon_radius": 0,
|
| 50 |
+
},
|
| 51 |
+
rtc_configuration=get_twilio_turn_credentials(),
|
| 52 |
+
additional_inputs=[
|
| 53 |
+
gr.Radio(
|
| 54 |
+
choices=["moonshine/base", "moonshine/tiny"],
|
| 55 |
+
value="moonshine/base",
|
| 56 |
+
label="Model",
|
| 57 |
+
)
|
| 58 |
+
],
|
| 59 |
+
additional_outputs=[gr.Textbox(label="Captions")],
|
| 60 |
+
additional_outputs_handler=lambda prev, current: (prev + "\n" + current).strip(),
|
| 61 |
+
)
|
| 62 |
+
|
| 63 |
+
if __name__ == "__main__":
|
| 64 |
+
stream.ui.launch()
|
default-favicon.ico
ADDED
|
|
requirements.txt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
fastrtc[vad]
|
| 2 |
+
useful-moonshine-onnx@git+https://[email protected]/usefulsensors/moonshine.git#subdirectory=moonshine-onnx
|
| 3 |
+
twilio
|