Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -24,67 +24,38 @@ hf_token = os.getenv("HF_TOKEN")
|
|
24 |
if hf_token:
|
25 |
login(token=hf_token)
|
26 |
|
27 |
-
# Global variables for model caching
|
28 |
_tts_model = None
|
29 |
_speakers_dict = None
|
30 |
_model_initialized = False
|
31 |
_initialization_in_progress = False
|
32 |
|
33 |
def get_speakers_dict():
|
34 |
-
"""Get speakers dictionary using the correct
|
35 |
try:
|
36 |
-
#
|
37 |
from maliba_ai.config.settings import Speakers
|
38 |
|
39 |
-
#
|
40 |
-
|
41 |
-
|
42 |
-
"
|
43 |
-
"
|
44 |
-
"
|
45 |
-
"
|
46 |
-
"
|
47 |
-
"
|
48 |
-
"
|
49 |
-
"
|
50 |
-
"Ibrahima":
|
51 |
-
"Amara": "Amara"
|
52 |
}
|
53 |
|
54 |
-
|
55 |
-
|
56 |
-
if hasattr(Speakers, attr_name):
|
57 |
-
available_speakers[name] = getattr(Speakers, attr_name)
|
58 |
-
except:
|
59 |
-
continue
|
60 |
|
61 |
-
if available_speakers:
|
62 |
-
logger.info(f"Loaded {len(available_speakers)} speakers from new structure: {list(available_speakers.keys())}")
|
63 |
-
return available_speakers
|
64 |
-
else:
|
65 |
-
raise AttributeError("No speakers found in new structure")
|
66 |
-
|
67 |
except Exception as e:
|
68 |
-
logger.error(f"Failed to import
|
69 |
-
|
70 |
-
try:
|
71 |
-
from maliba_ai.config.speakers import Adame, Moussa, Bourama, Modibo, Seydou, Amadou, Bakary, Ngolo, Ibrahima, Amara
|
72 |
-
logger.info("Using fallback old speaker structure")
|
73 |
-
return {
|
74 |
-
"Adama": Adame,
|
75 |
-
"Moussa": Moussa,
|
76 |
-
"Bourama": Bourama,
|
77 |
-
"Modibo": Modibo,
|
78 |
-
"Seydou": Seydou,
|
79 |
-
"Amadou": Amadou,
|
80 |
-
"Bakary": Bakary,
|
81 |
-
"Ngolo": Ngolo,
|
82 |
-
"Ibrahima": Ibrahima,
|
83 |
-
"Amara": Amara
|
84 |
-
}
|
85 |
-
except Exception as e2:
|
86 |
-
logger.error(f"Failed to import speakers: {e2}")
|
87 |
-
return {}
|
88 |
|
89 |
@spaces.GPU()
|
90 |
def initialize_model_once():
|
@@ -108,8 +79,8 @@ def initialize_model_once():
|
|
108 |
logger.info("Initializing Bambara TTS model...")
|
109 |
start_time = time.time()
|
110 |
|
111 |
-
# Use the
|
112 |
-
from maliba_ai.tts import BambaraTTSInference
|
113 |
|
114 |
model = BambaraTTSInference()
|
115 |
speakers = get_speakers_dict()
|
@@ -288,9 +259,9 @@ def build_interface():
|
|
288 |
gr.Markdown("""
|
289 |
# π€ Bambara Text-to-Speech
|
290 |
|
291 |
-
**Powered by MALIBA-AI**
|
292 |
|
293 |
-
Convert Bambara text to speech using our state-of-the-art TTS
|
294 |
|
295 |
**Bambara** is spoken by millions of people in Mali and West Africa.
|
296 |
""")
|
@@ -363,7 +334,7 @@ def build_interface():
|
|
363 |
label="Generated Speech",
|
364 |
type="numpy",
|
365 |
interactive=False,
|
366 |
-
format="wav"
|
367 |
)
|
368 |
|
369 |
status_output = gr.Textbox(
|
@@ -437,8 +408,7 @@ def build_interface():
|
|
437 |
def main():
|
438 |
"""Main function to launch the Gradio interface"""
|
439 |
logger.info("Starting Bambara TTS Gradio interface.")
|
440 |
-
|
441 |
-
# DO NOT preload - let it initialize on first request only (like your working version)
|
442 |
interface = build_interface()
|
443 |
interface.launch(
|
444 |
server_name="0.0.0.0",
|
|
|
24 |
if hf_token:
|
25 |
login(token=hf_token)
|
26 |
|
27 |
+
# Global variables for model caching
|
28 |
_tts_model = None
|
29 |
_speakers_dict = None
|
30 |
_model_initialized = False
|
31 |
_initialization_in_progress = False
|
32 |
|
33 |
def get_speakers_dict():
|
34 |
+
"""Get speakers dictionary using the correct SDK structure"""
|
35 |
try:
|
36 |
+
# Import the Speakers class (not individual speakers)
|
37 |
from maliba_ai.config.settings import Speakers
|
38 |
|
39 |
+
# Access all 10 speakers through the Speakers class
|
40 |
+
speakers_dict = {
|
41 |
+
"Adama": Speakers.Adama,
|
42 |
+
"Moussa": Speakers.Moussa,
|
43 |
+
"Bourama": Speakers.Bourama,
|
44 |
+
"Modibo": Speakers.Modibo,
|
45 |
+
"Seydou": Speakers.Seydou,
|
46 |
+
"Amadou": Speakers.Amadou,
|
47 |
+
"Bakary": Speakers.Bakary,
|
48 |
+
"Ngolo": Speakers.Ngolo,
|
49 |
+
"Amara": Speakers.Amara,
|
50 |
+
"Ibrahima": Speakers.Ibrahima
|
|
|
51 |
}
|
52 |
|
53 |
+
logger.info(f"π€ Successfully loaded {len(speakers_dict)} speakers: {list(speakers_dict.keys())}")
|
54 |
+
return speakers_dict
|
|
|
|
|
|
|
|
|
55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
except Exception as e:
|
57 |
+
logger.error(f"β Failed to import Speakers class: {e}")
|
58 |
+
return {}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
|
60 |
@spaces.GPU()
|
61 |
def initialize_model_once():
|
|
|
79 |
logger.info("Initializing Bambara TTS model...")
|
80 |
start_time = time.time()
|
81 |
|
82 |
+
# Use the correct import path
|
83 |
+
from maliba_ai.tts.inference import BambaraTTSInference
|
84 |
|
85 |
model = BambaraTTSInference()
|
86 |
speakers = get_speakers_dict()
|
|
|
259 |
gr.Markdown("""
|
260 |
# π€ Bambara Text-to-Speech
|
261 |
|
262 |
+
**Powered by MALIBA-AI** | *First Open-Source Bambara TTS*
|
263 |
|
264 |
+
Convert Bambara text to natural-sounding speech using our state-of-the-art neural TTS system.
|
265 |
|
266 |
**Bambara** is spoken by millions of people in Mali and West Africa.
|
267 |
""")
|
|
|
334 |
label="Generated Speech",
|
335 |
type="numpy",
|
336 |
interactive=False,
|
337 |
+
format="wav"
|
338 |
)
|
339 |
|
340 |
status_output = gr.Textbox(
|
|
|
408 |
def main():
|
409 |
"""Main function to launch the Gradio interface"""
|
410 |
logger.info("Starting Bambara TTS Gradio interface.")
|
411 |
+
|
|
|
412 |
interface = build_interface()
|
413 |
interface.launch(
|
414 |
server_name="0.0.0.0",
|