Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,40 +1,33 @@
|
|
|
|
|
|
|
|
1 |
import tempfile
|
2 |
import gradio as gr
|
|
|
|
|
3 |
from neon_tts_plugin_coqui import CoquiTTS
|
|
|
|
|
|
|
4 |
LANGUAGES = list(CoquiTTS.langs.keys())
|
5 |
LANGUAGES = LANGUAGES + ['cn', 'jp']
|
6 |
default_lang = "en"
|
7 |
-
#import whisper
|
8 |
-
#whisper_model = whisper.load_model("small")
|
9 |
-
#whisper = gr.Interface.load(name="spaces/abidlabs/whisper-large-v2")
|
10 |
whisper = gr.Interface.load(name="spaces/sanchit-gandhi/whisper-large-v2")
|
11 |
-
#cn_a_jp = gr.Blocks.load(name="spaces/Yusin/anime-tts_yusin")
|
12 |
-
#chatgpt = gr.Blocks.load(name="spaces/fffiloni/whisper-to-chatGPT")
|
13 |
-
#chatgpt = gr.Blocks.load(name="spaces/seawolf2357/chatgptclone")
|
14 |
-
import os
|
15 |
-
import json
|
16 |
-
import openai
|
17 |
-
#session_token = os.environ.get('SessionToken')
|
18 |
api_key = os.environ.get('api_key')
|
19 |
#if you have OpenAI API key as a string, enable the below
|
20 |
openai.api_key = api_key
|
21 |
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
coquiTTS = CoquiTTS()
|
26 |
-
|
27 |
|
28 |
# ChatGPT
|
29 |
-
def chat_hf(audio,
|
30 |
try:
|
31 |
whisper_text = translate(audio)
|
32 |
if whisper_text == "ERROR: You have to either use the microphone or upload an audio file":
|
33 |
gpt_response = "MISSING AUDIO: Record your voice by clicking the microphone button, do not forget to stop recording before sending your message ;)"
|
34 |
else:
|
35 |
-
#gpt_response = chatgpt(whisper_text, [], fn_index=0)
|
36 |
-
#print(gpt_response)
|
37 |
-
#gpt_response = gpt_response[0]
|
38 |
gpt_response = openai_create(whisper_text)
|
39 |
|
40 |
except:
|
@@ -42,35 +35,16 @@ def chat_hf(audio, custom_token, language):
|
|
42 |
gpt_response = """Sorry, I'm quite busy right now, but please try again later :)"""
|
43 |
|
44 |
# to voice
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
|
|
|
|
49 |
|
50 |
-
# whisper
|
51 |
-
#def translate(audio):
|
52 |
-
# print("""
|
53 |
-
# β
|
54 |
-
# Sending audio to Whisper ...
|
55 |
-
# β
|
56 |
-
# """)
|
57 |
-
#
|
58 |
-
# audio = whisper.load_audio(audio)
|
59 |
-
# audio = whisper.pad_or_trim(audio)
|
60 |
-
#
|
61 |
-
# mel = whisper.log_mel_spectrogram(audio).to(whisper_model.device)
|
62 |
-
#
|
63 |
-
# _, probs = whisper_model.detect_language(mel)
|
64 |
-
#
|
65 |
-
# transcript_options = whisper.DecodingOptions(task="transcribe", fp16 = False)
|
66 |
-
#
|
67 |
-
# transcription = whisper.decode(whisper_model, mel, transcript_options)
|
68 |
-
#
|
69 |
-
# print("language spoken: " + transcription.language)
|
70 |
-
# print("transcript: " + transcription.text)
|
71 |
-
# print("βββββββββββββββββββββββββββββββββββββββββββ")
|
72 |
-
#
|
73 |
-
# return transcription.text
|
74 |
|
75 |
def translate(audio):
|
76 |
print("""
|
@@ -78,14 +52,13 @@ def translate(audio):
|
|
78 |
Sending audio to Whisper ...
|
79 |
β
|
80 |
""")
|
81 |
-
|
82 |
text_result = whisper(audio, None, "transcribe", fn_index=0)
|
83 |
print(text_result)
|
84 |
return text_result
|
85 |
|
86 |
|
87 |
def openai_create(prompt):
|
88 |
-
|
89 |
response = openai.Completion.create(
|
90 |
model="text-davinci-003",
|
91 |
prompt=prompt,
|
@@ -99,11 +72,9 @@ def openai_create(prompt):
|
|
99 |
print(response.choices[0].text)
|
100 |
return response.choices[0].text
|
101 |
|
|
|
102 |
with gr.Blocks() as blocks:
|
103 |
-
gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>"
|
104 |
-
+ title
|
105 |
-
+ "</h1>")
|
106 |
-
#gr.Markdown(description)
|
107 |
radio = gr.Radio(label="Language", choices=LANGUAGES, value=default_lang)
|
108 |
with gr.Row(equal_height=True):# equal_height=False
|
109 |
with gr.Column():# variant="panel"
|
@@ -115,18 +86,11 @@ with gr.Blocks() as blocks:
|
|
115 |
text1 = gr.Textbox(label="Speech to Text")
|
116 |
text2 = gr.Textbox(label="ChatGPT Response")
|
117 |
audio = gr.Audio(label="Output", interactive=False)
|
118 |
-
#gr.Markdown(info)
|
119 |
-
#gr.Markdown("<center>"
|
120 |
-
# +f'<img src={badge} alt="visitors badge"/>'
|
121 |
-
# +"</center>")
|
122 |
-
|
123 |
# actions
|
124 |
submit.click(
|
125 |
chat_hf,
|
126 |
[audio_file, custom_token, radio],
|
127 |
[text1, text2, audio],
|
128 |
)
|
129 |
-
#radio.change(lambda lang: CoquiTTS.langs[lang]["sentence"], radio, text2)
|
130 |
-
|
131 |
|
132 |
blocks.launch(debug=True)
|
|
|
1 |
+
import os
|
2 |
+
import json
|
3 |
+
import openai
|
4 |
import tempfile
|
5 |
import gradio as gr
|
6 |
+
import infer
|
7 |
+
import config
|
8 |
from neon_tts_plugin_coqui import CoquiTTS
|
9 |
+
title = "Speech to ChatGPT to Speech"
|
10 |
+
coquiTTS = CoquiTTS()
|
11 |
+
|
12 |
LANGUAGES = list(CoquiTTS.langs.keys())
|
13 |
LANGUAGES = LANGUAGES + ['cn', 'jp']
|
14 |
default_lang = "en"
|
|
|
|
|
|
|
15 |
whisper = gr.Interface.load(name="spaces/sanchit-gandhi/whisper-large-v2")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
api_key = os.environ.get('api_key')
|
17 |
#if you have OpenAI API key as a string, enable the below
|
18 |
openai.api_key = api_key
|
19 |
|
20 |
+
pth_path = config.pth_path
|
21 |
+
config_json = config.config_json
|
22 |
+
net_g_ms, hps = infer.load_model(config_json, pth_path)
|
|
|
|
|
23 |
|
24 |
# ChatGPT
|
25 |
+
def chat_hf(audio, language):
|
26 |
try:
|
27 |
whisper_text = translate(audio)
|
28 |
if whisper_text == "ERROR: You have to either use the microphone or upload an audio file":
|
29 |
gpt_response = "MISSING AUDIO: Record your voice by clicking the microphone button, do not forget to stop recording before sending your message ;)"
|
30 |
else:
|
|
|
|
|
|
|
31 |
gpt_response = openai_create(whisper_text)
|
32 |
|
33 |
except:
|
|
|
35 |
gpt_response = """Sorry, I'm quite busy right now, but please try again later :)"""
|
36 |
|
37 |
# to voice
|
38 |
+
if language == 'cn' or 'jp':
|
39 |
+
text = infer.clean_text(gpt_response)
|
40 |
+
audio = infer.infer(text, net_g_ms, 2, "demo")
|
41 |
+
voice_out = (hps.data.sampling_rate, audio)
|
42 |
+
else:
|
43 |
+
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
|
44 |
+
coquiTTS.get_tts(gpt_response, fp, speaker = {"language" : language})
|
45 |
+
voice_out = fp.name
|
46 |
+
return whisper_text, gpt_response, voice_out
|
47 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
def translate(audio):
|
50 |
print("""
|
|
|
52 |
Sending audio to Whisper ...
|
53 |
β
|
54 |
""")
|
55 |
+
|
56 |
text_result = whisper(audio, None, "transcribe", fn_index=0)
|
57 |
print(text_result)
|
58 |
return text_result
|
59 |
|
60 |
|
61 |
def openai_create(prompt):
|
|
|
62 |
response = openai.Completion.create(
|
63 |
model="text-davinci-003",
|
64 |
prompt=prompt,
|
|
|
72 |
print(response.choices[0].text)
|
73 |
return response.choices[0].text
|
74 |
|
75 |
+
|
76 |
with gr.Blocks() as blocks:
|
77 |
+
gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>" + title + "</h1>")
|
|
|
|
|
|
|
78 |
radio = gr.Radio(label="Language", choices=LANGUAGES, value=default_lang)
|
79 |
with gr.Row(equal_height=True):# equal_height=False
|
80 |
with gr.Column():# variant="panel"
|
|
|
86 |
text1 = gr.Textbox(label="Speech to Text")
|
87 |
text2 = gr.Textbox(label="ChatGPT Response")
|
88 |
audio = gr.Audio(label="Output", interactive=False)
|
|
|
|
|
|
|
|
|
|
|
89 |
# actions
|
90 |
submit.click(
|
91 |
chat_hf,
|
92 |
[audio_file, custom_token, radio],
|
93 |
[text1, text2, audio],
|
94 |
)
|
|
|
|
|
95 |
|
96 |
blocks.launch(debug=True)
|