Update app.py
Browse files
app.py
CHANGED
|
@@ -7,17 +7,15 @@ import uuid
|
|
| 7 |
from elevenlabs import VoiceSettings
|
| 8 |
from elevenlabs.client import ElevenLabs
|
| 9 |
from pathlib import Path
|
| 10 |
-
from scipy.io.wavfile import write
|
| 11 |
-
from scipy.io.wavfile import read
|
| 12 |
import yt_dlp
|
| 13 |
|
| 14 |
ELEVENLABS_API = os.environ.get("ELEVENLABS_API")
|
| 15 |
-
|
| 16 |
ASSEMBLYAI_API = os.environ.get("ASSEMBLYAI_API")
|
| 17 |
|
| 18 |
def voice_to_voice(audio_file):
|
| 19 |
transcript = transcribe_audio(audio_file)
|
| 20 |
-
if transcript.status ==
|
| 21 |
raise gr.Error(transcript.error)
|
| 22 |
else:
|
| 23 |
transcript = transcript.text
|
|
@@ -33,7 +31,7 @@ def voice_to_voice(audio_file):
|
|
| 33 |
return tuple(generated_audio_paths + list_translations)
|
| 34 |
|
| 35 |
def transcribe_audio(audio_file):
|
| 36 |
-
aai.settings.api_key =
|
| 37 |
transcriber = aai.Transcriber()
|
| 38 |
transcript = transcriber.transcribe(audio_file)
|
| 39 |
return transcript
|
|
@@ -73,8 +71,6 @@ def text_to_speech(text):
|
|
| 73 |
|
| 74 |
return save_file_path
|
| 75 |
|
| 76 |
-
|
| 77 |
-
|
| 78 |
def download_audio(url):
|
| 79 |
ydl_opts = {
|
| 80 |
'format': 'bestaudio/best',
|
|
@@ -94,36 +90,34 @@ def download_audio(url):
|
|
| 94 |
|
| 95 |
return sample_rate, audio_array
|
| 96 |
|
| 97 |
-
|
| 98 |
with gr.Blocks() as demo:
|
| 99 |
-
gr.Markdown("##
|
| 100 |
gr.Markdown(
|
| 101 |
-
|
| 102 |
The API Key you need:
|
| 103 |
-
|
| 104 |
-
|
| 105 |
Note: you need at least 30 minutes of a voice recording of yourself for the *Professional voice cloning. But there is also a simpler voice cloning option that only requires 30 seconds of voice recording. *Professional voice cloning is a paid feature.
|
| 106 |
-
|
| 107 |
"""
|
| 108 |
)
|
| 109 |
audio_input = gr.Audio(type="filepath", show_download_button=True)
|
| 110 |
-
with gr.Accordion("
|
| 111 |
with gr.Row():
|
| 112 |
link = gr.Textbox(
|
| 113 |
-
label
|
| 114 |
-
placeholder
|
| 115 |
-
interactive
|
| 116 |
)
|
| 117 |
download_button = gr.Button(
|
| 118 |
"Download!",
|
| 119 |
-
variant
|
| 120 |
)
|
| 121 |
download_button.click(download_audio, [link], [audio_input])
|
| 122 |
submit = gr.Button("Submit", variant="primary")
|
| 123 |
clear_button = gr.ClearButton(audio_input, "Clear")
|
| 124 |
|
| 125 |
output_components = []
|
| 126 |
-
languages = ["Turkish", "Swedish", "Russian", "German", "Spanish", "Japanese", "
|
| 127 |
|
| 128 |
for lang in languages:
|
| 129 |
with gr.Group():
|
|
|
|
| 7 |
from elevenlabs import VoiceSettings
|
| 8 |
from elevenlabs.client import ElevenLabs
|
| 9 |
from pathlib import Path
|
| 10 |
+
from scipy.io.wavfile import write, read
|
|
|
|
| 11 |
import yt_dlp
|
| 12 |
|
| 13 |
ELEVENLABS_API = os.environ.get("ELEVENLABS_API")
|
|
|
|
| 14 |
ASSEMBLYAI_API = os.environ.get("ASSEMBLYAI_API")
|
| 15 |
|
| 16 |
def voice_to_voice(audio_file):
|
| 17 |
transcript = transcribe_audio(audio_file)
|
| 18 |
+
if transcript.status == 'error':
|
| 19 |
raise gr.Error(transcript.error)
|
| 20 |
else:
|
| 21 |
transcript = transcript.text
|
|
|
|
| 31 |
return tuple(generated_audio_paths + list_translations)
|
| 32 |
|
| 33 |
def transcribe_audio(audio_file):
|
| 34 |
+
aai.settings.api_key = ASSEMBLYAI_API
|
| 35 |
transcriber = aai.Transcriber()
|
| 36 |
transcript = transcriber.transcribe(audio_file)
|
| 37 |
return transcript
|
|
|
|
| 71 |
|
| 72 |
return save_file_path
|
| 73 |
|
|
|
|
|
|
|
| 74 |
def download_audio(url):
|
| 75 |
ydl_opts = {
|
| 76 |
'format': 'bestaudio/best',
|
|
|
|
| 90 |
|
| 91 |
return sample_rate, audio_array
|
| 92 |
|
|
|
|
| 93 |
with gr.Blocks() as demo:
|
| 94 |
+
gr.Markdown("## Audio Translator")
|
| 95 |
gr.Markdown(
|
| 96 |
+
"""
|
| 97 |
The API Key you need:
|
| 98 |
+
[AssemblyAI API key](https://www.assemblyai.com/?utm_source=youtube&utm_medium=referral&utm_campaign=yt_mis_66)<br>
|
| 99 |
+
[Elevenlabs API key](https://elevenlabs.io/)<br>
|
| 100 |
Note: you need at least 30 minutes of a voice recording of yourself for the *Professional voice cloning. But there is also a simpler voice cloning option that only requires 30 seconds of voice recording. *Professional voice cloning is a paid feature.
|
|
|
|
| 101 |
"""
|
| 102 |
)
|
| 103 |
audio_input = gr.Audio(type="filepath", show_download_button=True)
|
| 104 |
+
with gr.Accordion("Inputs by Link", open=False):
|
| 105 |
with gr.Row():
|
| 106 |
link = gr.Textbox(
|
| 107 |
+
label="Link",
|
| 108 |
+
placeholder="Paste the link here",
|
| 109 |
+
interactive=True
|
| 110 |
)
|
| 111 |
download_button = gr.Button(
|
| 112 |
"Download!",
|
| 113 |
+
variant="primary"
|
| 114 |
)
|
| 115 |
download_button.click(download_audio, [link], [audio_input])
|
| 116 |
submit = gr.Button("Submit", variant="primary")
|
| 117 |
clear_button = gr.ClearButton(audio_input, "Clear")
|
| 118 |
|
| 119 |
output_components = []
|
| 120 |
+
languages = ["Turkish", "Swedish", "Russian", "German", "Spanish", "Japanese", "Indonesian"]
|
| 121 |
|
| 122 |
for lang in languages:
|
| 123 |
with gr.Group():
|