Spaces:
Build error
Build error
ss
Browse files
app.py
CHANGED
@@ -15,14 +15,29 @@ HF_TOKEN = os.environ["HF_TOKEN"]
|
|
15 |
headers = {"Authorization": f"Bearer {HF_TOKEN}"}
|
16 |
#Language covered in Bloom : en, fr, esp, arb, hn, portu, Indonesian, Vietnamese, Chinese, tamil, telugu, bengali
|
17 |
|
|
|
|
|
|
|
|
|
18 |
# Text-to-Speech
|
19 |
LANGUAGES = list(CoquiTTS.langs.keys())
|
20 |
print(f"Languages for Coqui are: {LANGUAGES}")
|
21 |
#Languages for Coqui are: ['en', 'es', 'fr', 'de', 'pl', 'uk', 'ro', 'hu', 'el', 'bg', 'nl', 'fi', 'sl', 'lv', 'ga']
|
22 |
coquiTTS = CoquiTTS()
|
23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
24 |
# Whisper - speeech-to-text
|
25 |
def whisper_stt(audio):
|
|
|
26 |
# load audio and pad/trim it to fit 30 seconds
|
27 |
audio = whisper.load_audio(audio)
|
28 |
audio = whisper.pad_or_trim(audio)
|
@@ -41,19 +56,16 @@ def whisper_stt(audio):
|
|
41 |
|
42 |
# print the recognized text
|
43 |
print(f"transcript is : {result.text}")
|
44 |
-
return result.text
|
45 |
|
46 |
-
# Driver function
|
47 |
-
def fun_engine(audio) :
|
48 |
-
text1 = whisper_stt(audio)
|
49 |
-
#text1 = model.transcribe(audio)["text"]
|
50 |
-
text2 = lang_model_response(text1)
|
51 |
-
speech = tts(text2, 'en')
|
52 |
-
return text1, text2, speech
|
53 |
|
54 |
# LLM - Bloom Response
|
55 |
def lang_model_response(prompt):
|
56 |
print(f"*****Inside lang_model_response - Prompt is :{prompt}")
|
|
|
|
|
|
|
|
|
57 |
if len(prompt) == 0:
|
58 |
prompt = """Can you help me please?"""
|
59 |
|
@@ -82,6 +94,7 @@ def lang_model_response(prompt):
|
|
82 |
|
83 |
# Coqui - Text-to-Speech
|
84 |
def tts(text, language):
|
|
|
85 |
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
|
86 |
coquiTTS.get_tts(text, fp, speaker = {"language" : language})
|
87 |
return fp.name
|
|
|
15 |
headers = {"Authorization": f"Bearer {HF_TOKEN}"}
|
16 |
#Language covered in Bloom : en, fr, esp, arb, hn, portu, Indonesian, Vietnamese, Chinese, tamil, telugu, bengali
|
17 |
|
18 |
+
prompt = """Instruction: Given a Statement, produce a response in one sentence.
|
19 |
+
Statement:
|
20 |
+
"""
|
21 |
+
|
22 |
# Text-to-Speech
|
23 |
LANGUAGES = list(CoquiTTS.langs.keys())
|
24 |
print(f"Languages for Coqui are: {LANGUAGES}")
|
25 |
#Languages for Coqui are: ['en', 'es', 'fr', 'de', 'pl', 'uk', 'ro', 'hu', 'el', 'bg', 'nl', 'fi', 'sl', 'lv', 'ga']
|
26 |
coquiTTS = CoquiTTS()
|
27 |
|
28 |
+
|
29 |
+
# Driver function
|
30 |
+
def fun_engine(audio) :
|
31 |
+
text1, lang = whisper_stt(audio)
|
32 |
+
#text1 = model.transcribe(audio)["text"]
|
33 |
+
text2 = lang_model_response(text1)
|
34 |
+
speech = tts(text2, lang) #'en')
|
35 |
+
return text1, text2, speech
|
36 |
+
|
37 |
+
|
38 |
# Whisper - speeech-to-text
|
39 |
def whisper_stt(audio):
|
40 |
+
print("Inside Whisper TTS")
|
41 |
# load audio and pad/trim it to fit 30 seconds
|
42 |
audio = whisper.load_audio(audio)
|
43 |
audio = whisper.pad_or_trim(audio)
|
|
|
56 |
|
57 |
# print the recognized text
|
58 |
print(f"transcript is : {result.text}")
|
59 |
+
return result.text, lang
|
60 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
|
62 |
# LLM - Bloom Response
|
63 |
def lang_model_response(prompt):
|
64 |
print(f"*****Inside lang_model_response - Prompt is :{prompt}")
|
65 |
+
p = """Instruction: Given a Statement, produce a Response in one sentence.
|
66 |
+
Statement: """
|
67 |
+
prompt = p + prompt + "\n" + "Response: "
|
68 |
+
|
69 |
if len(prompt) == 0:
|
70 |
prompt = """Can you help me please?"""
|
71 |
|
|
|
94 |
|
95 |
# Coqui - Text-to-Speech
|
96 |
def tts(text, language):
|
97 |
+
print(f"Inside tts - language is : {language}")
|
98 |
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
|
99 |
coquiTTS.get_tts(text, fp, speaker = {"language" : language})
|
100 |
return fp.name
|