Spaces:
Paused
Paused
rick
commited on
minors changes
Browse files- core/core.py +2 -1
- core/speech_to_text.py +44 -0
core/core.py
CHANGED
|
@@ -107,4 +107,5 @@ def init_process_mode(
|
|
| 107 |
dest_lang=f"{to_lang}"
|
| 108 |
)
|
| 109 |
return system_prompt, operation_prompt
|
| 110 |
-
return "", ""
|
|
|
|
|
|
| 107 |
dest_lang=f"{to_lang}"
|
| 108 |
)
|
| 109 |
return system_prompt, operation_prompt
|
| 110 |
+
return "", ""
|
| 111 |
+
|
core/speech_to_text.py
CHANGED
|
@@ -114,6 +114,50 @@ def transcribe_audio(filepath: Union[str, IO], language: Optional[str] = None) -
|
|
| 114 |
return ""
|
| 115 |
|
| 116 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 117 |
|
| 118 |
|
| 119 |
|
|
|
|
| 114 |
return ""
|
| 115 |
|
| 116 |
|
| 117 |
+
# ############################################################
|
| 118 |
+
|
| 119 |
+
def translate_audio(filepath: Union[str, IO]) -> str:
|
| 120 |
+
"""
|
| 121 |
+
Traduit un fichier audio temporaire en Anglais.
|
| 122 |
+
|
| 123 |
+
Args:
|
| 124 |
+
filepath Chemin vers le fichier audio temporaire à traduire.
|
| 125 |
+
|
| 126 |
+
Returns:
|
| 127 |
+
str: Le texte traduit.
|
| 128 |
+
"""
|
| 129 |
+
translated_text = []
|
| 130 |
+
client = OpenAI(api_key=getenv("OPENAI_API_KEY"))
|
| 131 |
+
try:
|
| 132 |
+
with open(filepath if isinstance(filepath, str) else filepath.name, "rb") as f:
|
| 133 |
+
# filepath peut etre un chemin vers un fichier audio ou un objet IO
|
| 134 |
+
f.seek(0)
|
| 135 |
+
audio = AudioSegment.from_file(f)
|
| 136 |
+
duration_ms = len(audio)
|
| 137 |
+
segment_duration_ms = int(
|
| 138 |
+
(max_size_mb * 1024 * 1024 * 8) /
|
| 139 |
+
(audio.frame_rate * audio.sample_width * audio.channels)
|
| 140 |
+
)
|
| 141 |
+
|
| 142 |
+
for start in range(0, duration_ms, segment_duration_ms):
|
| 143 |
+
end = min(start + segment_duration_ms, duration_ms)
|
| 144 |
+
segment = audio[start:end]
|
| 145 |
+
|
| 146 |
+
buffer = BytesIO()
|
| 147 |
+
segment.export(buffer, format="mp3")
|
| 148 |
+
buffer.seek(0)
|
| 149 |
+
|
| 150 |
+
translation = client.audio.translations.create(
|
| 151 |
+
model="whisper-1",
|
| 152 |
+
file=audio_file
|
| 153 |
+
)
|
| 154 |
+
translated_text.append(translation)
|
| 155 |
+
|
| 156 |
+
return " ".join(translated_text)
|
| 157 |
+
except Exception as e:
|
| 158 |
+
print(f"Erreur lors de la traduction de l'audio : {e}")
|
| 159 |
+
return ""
|
| 160 |
+
|
| 161 |
|
| 162 |
|
| 163 |
|