Spaces:
Sleeping
Sleeping
Commit
·
7089c67
1
Parent(s):
ae07559
chore: Update deps
Browse files- app.py +5 -5
- requirements.txt +21 -14
app.py
CHANGED
|
@@ -1,7 +1,6 @@
|
|
| 1 |
"""Røst speech-to-text demo."""
|
| 2 |
|
| 3 |
import logging
|
| 4 |
-
import os
|
| 5 |
import warnings
|
| 6 |
|
| 7 |
import gradio as gr
|
|
@@ -11,6 +10,7 @@ import torch
|
|
| 11 |
from punctfix import PunctFixer
|
| 12 |
from transformers import pipeline
|
| 13 |
from dotenv import load_dotenv
|
|
|
|
| 14 |
|
| 15 |
logging.basicConfig(
|
| 16 |
level=logging.INFO,
|
|
@@ -69,6 +69,8 @@ transcriber = pipeline(
|
|
| 69 |
logger.info("Loading the punctuation fixer model...")
|
| 70 |
transcription_fixer = PunctFixer(language="da", device=device)
|
| 71 |
|
|
|
|
|
|
|
| 72 |
logger.info("Models loaded, ready to transcribe audio.")
|
| 73 |
|
| 74 |
def transcribe_audio(sampling_rate_and_audio: tuple[int, np.ndarray] | None) -> str:
|
|
@@ -92,6 +94,7 @@ def transcribe_audio(sampling_rate_and_audio: tuple[int, np.ndarray] | None) ->
|
|
| 92 |
if audio.ndim > 1:
|
| 93 |
audio = np.mean(audio, axis=1)
|
| 94 |
audio = samplerate.resample(audio, 16_000 / sampling_rate, "sinc_best")
|
|
|
|
| 95 |
|
| 96 |
logger.info(f"Transcribing audio clip of {len(audio) / 16_000:.2f} seconds...")
|
| 97 |
transcription = transcriber(
|
|
@@ -111,13 +114,11 @@ def transcribe_audio(sampling_rate_and_audio: tuple[int, np.ndarray] | None) ->
|
|
| 111 |
demo = gr.Interface(
|
| 112 |
fn=transcribe_audio,
|
| 113 |
inputs=gr.Audio(
|
| 114 |
-
sources=["microphone", "upload"], show_label=False
|
| 115 |
),
|
| 116 |
outputs="textbox",
|
| 117 |
title=TITLE,
|
| 118 |
description=DESCRIPTION,
|
| 119 |
-
css="p { font-size: 1.0rem; }",
|
| 120 |
-
allow_flagging="never",
|
| 121 |
examples=[
|
| 122 |
"https://filedn.com/lRBwPhPxgV74tO0rDoe8SpH/audio-examples/bornholmsk.wav",
|
| 123 |
"https://filedn.com/lRBwPhPxgV74tO0rDoe8SpH/audio-examples/soenderjysk.wav",
|
|
@@ -125,7 +126,6 @@ demo = gr.Interface(
|
|
| 125 |
"https://filedn.com/lRBwPhPxgV74tO0rDoe8SpH/audio-examples/accent.wav",
|
| 126 |
],
|
| 127 |
cache_examples=False,
|
| 128 |
-
theme=gr.themes.Soft(primary_hue="orange"),
|
| 129 |
)
|
| 130 |
|
| 131 |
demo.launch()
|
|
|
|
| 1 |
"""Røst speech-to-text demo."""
|
| 2 |
|
| 3 |
import logging
|
|
|
|
| 4 |
import warnings
|
| 5 |
|
| 6 |
import gradio as gr
|
|
|
|
| 10 |
from punctfix import PunctFixer
|
| 11 |
from transformers import pipeline
|
| 12 |
from dotenv import load_dotenv
|
| 13 |
+
import torch_audiomentations as ta
|
| 14 |
|
| 15 |
logging.basicConfig(
|
| 16 |
level=logging.INFO,
|
|
|
|
| 69 |
logger.info("Loading the punctuation fixer model...")
|
| 70 |
transcription_fixer = PunctFixer(language="da", device=device)
|
| 71 |
|
| 72 |
+
normaliser = ta.PeakNormalization(p=1.0)
|
| 73 |
+
|
| 74 |
logger.info("Models loaded, ready to transcribe audio.")
|
| 75 |
|
| 76 |
def transcribe_audio(sampling_rate_and_audio: tuple[int, np.ndarray] | None) -> str:
|
|
|
|
| 94 |
if audio.ndim > 1:
|
| 95 |
audio = np.mean(audio, axis=1)
|
| 96 |
audio = samplerate.resample(audio, 16_000 / sampling_rate, "sinc_best")
|
| 97 |
+
audio = normaliser(torch.tensor(audio).unsqueeze(0).unsqueeze(0)).squeeze().numpy()
|
| 98 |
|
| 99 |
logger.info(f"Transcribing audio clip of {len(audio) / 16_000:.2f} seconds...")
|
| 100 |
transcription = transcriber(
|
|
|
|
| 114 |
demo = gr.Interface(
|
| 115 |
fn=transcribe_audio,
|
| 116 |
inputs=gr.Audio(
|
| 117 |
+
sources=["microphone", "upload"], show_label=False
|
| 118 |
),
|
| 119 |
outputs="textbox",
|
| 120 |
title=TITLE,
|
| 121 |
description=DESCRIPTION,
|
|
|
|
|
|
|
| 122 |
examples=[
|
| 123 |
"https://filedn.com/lRBwPhPxgV74tO0rDoe8SpH/audio-examples/bornholmsk.wav",
|
| 124 |
"https://filedn.com/lRBwPhPxgV74tO0rDoe8SpH/audio-examples/soenderjysk.wav",
|
|
|
|
| 126 |
"https://filedn.com/lRBwPhPxgV74tO0rDoe8SpH/audio-examples/accent.wav",
|
| 127 |
],
|
| 128 |
cache_examples=False,
|
|
|
|
| 129 |
)
|
| 130 |
|
| 131 |
demo.launch()
|
requirements.txt
CHANGED
|
@@ -1,29 +1,33 @@
|
|
| 1 |
aiofiles==23.2.1
|
|
|
|
| 2 |
annotated-types==0.7.0
|
| 3 |
anyio==4.4.0
|
| 4 |
attrs==24.2.0
|
|
|
|
| 5 |
certifi==2024.8.30
|
| 6 |
charset-normalizer==3.3.2
|
| 7 |
click==8.1.7
|
| 8 |
contourpy==1.3.0
|
| 9 |
cycler==0.12.1
|
| 10 |
exceptiongroup==1.2.2
|
| 11 |
-
fastapi==0.
|
| 12 |
ffmpy==0.4.0
|
| 13 |
filelock==3.16.1
|
| 14 |
fonttools==4.53.1
|
| 15 |
fsspec==2024.9.0
|
| 16 |
-
gradio==
|
| 17 |
-
gradio_client==
|
|
|
|
| 18 |
h11==0.14.0
|
|
|
|
| 19 |
httpcore==1.0.5
|
| 20 |
httpx==0.27.2
|
| 21 |
-
huggingface-hub==0.
|
| 22 |
hypothesis==6.112.1
|
| 23 |
idna==3.10
|
| 24 |
importlib_resources==6.4.5
|
| 25 |
Jinja2==3.1.4
|
| 26 |
-
kenlm @ https://github.com/kpu/kenlm/archive/master.zip
|
| 27 |
kiwisolver==1.4.7
|
| 28 |
markdown-it-py==3.0.0
|
| 29 |
MarkupSafe==2.1.5
|
|
@@ -38,21 +42,22 @@ pandas==2.2.2
|
|
| 38 |
pillow==10.4.0
|
| 39 |
punctfix==0.11.1
|
| 40 |
pyctcdecode==0.5.0
|
| 41 |
-
pydantic==2.
|
| 42 |
-
pydantic_core==2.
|
| 43 |
pydub==0.25.1
|
| 44 |
Pygments==2.18.0
|
| 45 |
pygtrie==2.5.0
|
| 46 |
pyparsing==3.1.4
|
| 47 |
python-dateutil==2.9.0.post0
|
| 48 |
python-dotenv==1.0.1
|
| 49 |
-
python-multipart==0.0.
|
| 50 |
pytz==2024.2
|
| 51 |
PyYAML==6.0.2
|
| 52 |
regex==2024.9.11
|
| 53 |
requests==2.32.3
|
| 54 |
rich==13.8.1
|
| 55 |
ruff==0.6.5
|
|
|
|
| 56 |
safetensors==0.4.5
|
| 57 |
samplerate==0.2.1
|
| 58 |
semantic-version==2.10.0
|
|
@@ -60,15 +65,17 @@ shellingham==1.5.4
|
|
| 60 |
six==1.16.0
|
| 61 |
sniffio==1.3.1
|
| 62 |
sortedcontainers==2.4.0
|
| 63 |
-
starlette==0.
|
| 64 |
-
sympy==1.
|
| 65 |
-
tokenizers==0.
|
| 66 |
tomlkit==0.12.0
|
| 67 |
-
torch==2.
|
| 68 |
tqdm==4.66.5
|
| 69 |
-
transformers==4.
|
| 70 |
typer==0.12.5
|
| 71 |
-
|
|
|
|
|
|
|
| 72 |
tzdata==2024.1
|
| 73 |
urllib3==2.2.3
|
| 74 |
uvicorn==0.30.6
|
|
|
|
| 1 |
aiofiles==23.2.1
|
| 2 |
+
annotated-doc==0.0.4
|
| 3 |
annotated-types==0.7.0
|
| 4 |
anyio==4.4.0
|
| 5 |
attrs==24.2.0
|
| 6 |
+
brotli==1.2.0
|
| 7 |
certifi==2024.8.30
|
| 8 |
charset-normalizer==3.3.2
|
| 9 |
click==8.1.7
|
| 10 |
contourpy==1.3.0
|
| 11 |
cycler==0.12.1
|
| 12 |
exceptiongroup==1.2.2
|
| 13 |
+
fastapi==0.122.0
|
| 14 |
ffmpy==0.4.0
|
| 15 |
filelock==3.16.1
|
| 16 |
fonttools==4.53.1
|
| 17 |
fsspec==2024.9.0
|
| 18 |
+
gradio==6.0.1
|
| 19 |
+
gradio_client==2.0.0
|
| 20 |
+
groovy==0.1.2
|
| 21 |
h11==0.14.0
|
| 22 |
+
hf-xet==1.2.0
|
| 23 |
httpcore==1.0.5
|
| 24 |
httpx==0.27.2
|
| 25 |
+
huggingface-hub==0.36.0
|
| 26 |
hypothesis==6.112.1
|
| 27 |
idna==3.10
|
| 28 |
importlib_resources==6.4.5
|
| 29 |
Jinja2==3.1.4
|
| 30 |
+
kenlm @ https://github.com/kpu/kenlm/archive/master.zip#sha256=d23d300d559a45a5e3ede958dbbf2395231119c0b8cd97a1ea43480625894ff4
|
| 31 |
kiwisolver==1.4.7
|
| 32 |
markdown-it-py==3.0.0
|
| 33 |
MarkupSafe==2.1.5
|
|
|
|
| 42 |
pillow==10.4.0
|
| 43 |
punctfix==0.11.1
|
| 44 |
pyctcdecode==0.5.0
|
| 45 |
+
pydantic==2.12.4
|
| 46 |
+
pydantic_core==2.41.5
|
| 47 |
pydub==0.25.1
|
| 48 |
Pygments==2.18.0
|
| 49 |
pygtrie==2.5.0
|
| 50 |
pyparsing==3.1.4
|
| 51 |
python-dateutil==2.9.0.post0
|
| 52 |
python-dotenv==1.0.1
|
| 53 |
+
python-multipart==0.0.20
|
| 54 |
pytz==2024.2
|
| 55 |
PyYAML==6.0.2
|
| 56 |
regex==2024.9.11
|
| 57 |
requests==2.32.3
|
| 58 |
rich==13.8.1
|
| 59 |
ruff==0.6.5
|
| 60 |
+
safehttpx==0.1.7
|
| 61 |
safetensors==0.4.5
|
| 62 |
samplerate==0.2.1
|
| 63 |
semantic-version==2.10.0
|
|
|
|
| 65 |
six==1.16.0
|
| 66 |
sniffio==1.3.1
|
| 67 |
sortedcontainers==2.4.0
|
| 68 |
+
starlette==0.50.0
|
| 69 |
+
sympy==1.14.0
|
| 70 |
+
tokenizers==0.22.1
|
| 71 |
tomlkit==0.12.0
|
| 72 |
+
torch==2.9.1
|
| 73 |
tqdm==4.66.5
|
| 74 |
+
transformers==4.57.3
|
| 75 |
typer==0.12.5
|
| 76 |
+
typer-slim==0.20.0
|
| 77 |
+
typing-inspection==0.4.2
|
| 78 |
+
typing_extensions==4.15.0
|
| 79 |
tzdata==2024.1
|
| 80 |
urllib3==2.2.3
|
| 81 |
uvicorn==0.30.6
|