context and context score added
Browse files
app.py
CHANGED
|
@@ -29,6 +29,23 @@ SENTENCES = [
|
|
| 29 |
"Les systèmes vocaux progressent rapidement"
|
| 30 |
]
|
| 31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
# ============================
|
| 33 |
# CONFIGURATION
|
| 34 |
# ============================
|
|
@@ -77,7 +94,9 @@ f5tts = F5TTS()
|
|
| 77 |
|
| 78 |
print("F5-TTS loaded")
|
| 79 |
|
| 80 |
-
def generate_audio(model_name,
|
|
|
|
|
|
|
| 81 |
|
| 82 |
if model_name == "OmniVoice":
|
| 83 |
return generate_omnivoice(sentence)
|
|
@@ -146,8 +165,9 @@ def submit_annotation_hf(
|
|
| 146 |
tts_experience,
|
| 147 |
device_type,
|
| 148 |
selected_model,
|
| 149 |
-
|
| 150 |
-
score
|
|
|
|
| 151 |
):
|
| 152 |
annotation = {
|
| 153 |
"user_id": user_id,
|
|
@@ -158,9 +178,10 @@ def submit_annotation_hf(
|
|
| 158 |
"device_type": device_type,
|
| 159 |
|
| 160 |
"model_name": selected_model,
|
| 161 |
-
"
|
| 162 |
|
| 163 |
"score": score,
|
|
|
|
| 164 |
"timestamp": datetime.now(UTC).isoformat()
|
| 165 |
}
|
| 166 |
|
|
@@ -168,17 +189,17 @@ def submit_annotation_hf(
|
|
| 168 |
short_id = str(uuid.uuid4()) [:8]
|
| 169 |
|
| 170 |
file_name = (
|
| 171 |
-
f"
|
| 172 |
)
|
| 173 |
|
| 174 |
-
local_file = file_name
|
| 175 |
|
| 176 |
with open(local_file, "w", encoding="utf-8") as f:
|
| 177 |
json.dump(annotation, f, indent=2)
|
| 178 |
|
| 179 |
api.upload_file(
|
| 180 |
path_or_fileobj=local_file,
|
| 181 |
-
path_in_repo=f"annotations/{
|
| 182 |
repo_id=HF_DATASET_NAME,
|
| 183 |
repo_type="dataset"
|
| 184 |
)
|
|
@@ -194,10 +215,12 @@ def submit_annotation(
|
|
| 194 |
tts_experience,
|
| 195 |
device_type,
|
| 196 |
selected_model,
|
| 197 |
-
|
| 198 |
-
score
|
|
|
|
| 199 |
):
|
| 200 |
-
|
|
|
|
| 201 |
submit_annotation_hf(
|
| 202 |
user_id,
|
| 203 |
age_group,
|
|
@@ -206,11 +229,24 @@ def submit_annotation(
|
|
| 206 |
tts_experience,
|
| 207 |
device_type,
|
| 208 |
selected_model,
|
| 209 |
-
|
| 210 |
-
score
|
|
|
|
| 211 |
)
|
| 212 |
|
| 213 |
-
return f"Annotation saved for {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 214 |
|
| 215 |
# ============================
|
| 216 |
# SUBMIT BUTTON CONTROL
|
|
@@ -272,12 +308,12 @@ with gr.Blocks() as demo:
|
|
| 272 |
with gr.Row():
|
| 273 |
age_group = gr.Dropdown(
|
| 274 |
choices=[
|
| 275 |
-
"10-
|
| 276 |
-
"20-
|
| 277 |
-
"30-
|
| 278 |
-
"40-
|
| 279 |
-
"50-
|
| 280 |
-
"60-
|
| 281 |
"70+"
|
| 282 |
],
|
| 283 |
label="Age Group"
|
|
@@ -309,10 +345,15 @@ with gr.Blocks() as demo:
|
|
| 309 |
)
|
| 310 |
|
| 311 |
with gr.Row():
|
| 312 |
-
|
| 313 |
-
choices=
|
| 314 |
-
value=
|
| 315 |
-
label="
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 316 |
)
|
| 317 |
|
| 318 |
model_selector = gr.Dropdown(
|
|
@@ -321,6 +362,12 @@ with gr.Blocks() as demo:
|
|
| 321 |
label = "Model"
|
| 322 |
)
|
| 323 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 324 |
generate_btn = gr.Button("Generate Audio")
|
| 325 |
|
| 326 |
audio_player = gr.Audio(label="Generated audio")
|
|
@@ -329,13 +376,14 @@ with gr.Blocks() as demo:
|
|
| 329 |
generate_audio,
|
| 330 |
inputs=[
|
| 331 |
model_selector,
|
| 332 |
-
|
| 333 |
],
|
| 334 |
outputs=[audio_player, audio_played]
|
| 335 |
)
|
| 336 |
|
| 337 |
with gr.Row():
|
| 338 |
score = gr.Dropdown(choices=["None","1", "2", "3", "4", "5"], value="None", label="MOS Score (1–5)")
|
|
|
|
| 339 |
submit_btn = gr.Button("Submit Score", interactive=False)
|
| 340 |
|
| 341 |
status = gr.Textbox(label="Status", interactive=False)
|
|
@@ -375,7 +423,8 @@ with gr.Blocks() as demo:
|
|
| 375 |
native_language,
|
| 376 |
tts_experience,
|
| 377 |
device_type,
|
| 378 |
-
|
|
|
|
| 379 |
score
|
| 380 |
],
|
| 381 |
outputs=status
|
|
|
|
| 29 |
"Les systèmes vocaux progressent rapidement"
|
| 30 |
]
|
| 31 |
|
| 32 |
+
CONTEXTS = {
|
| 33 |
+
"Train station announcement":
|
| 34 |
+
"Le train 7891 à destination de Paris Nord, prévu à 14h35, partira voie 8.",
|
| 35 |
+
|
| 36 |
+
"Weather forecast":
|
| 37 |
+
"Demain, les températures atteindront 27 degrés avec un ciel dégagé.",
|
| 38 |
+
|
| 39 |
+
"Voice assistant":
|
| 40 |
+
"Je peux vous aider à trouver le restaurant le plus proche.",
|
| 41 |
+
|
| 42 |
+
"Audiobook":
|
| 43 |
+
"Le vieux château se dressait au sommet de la colline depuis plusieurs siècles.",
|
| 44 |
+
|
| 45 |
+
"Customer service":
|
| 46 |
+
"Votre demande a bien été prise en compte et sera traitée sous quarante-huit heures."
|
| 47 |
+
}
|
| 48 |
+
|
| 49 |
# ============================
|
| 50 |
# CONFIGURATION
|
| 51 |
# ============================
|
|
|
|
| 94 |
|
| 95 |
print("F5-TTS loaded")
|
| 96 |
|
| 97 |
+
def generate_audio(model_name, context):
|
| 98 |
+
|
| 99 |
+
sentence = CONTEXTS[context]
|
| 100 |
|
| 101 |
if model_name == "OmniVoice":
|
| 102 |
return generate_omnivoice(sentence)
|
|
|
|
| 165 |
tts_experience,
|
| 166 |
device_type,
|
| 167 |
selected_model,
|
| 168 |
+
context,
|
| 169 |
+
score,
|
| 170 |
+
context_score
|
| 171 |
):
|
| 172 |
annotation = {
|
| 173 |
"user_id": user_id,
|
|
|
|
| 178 |
"device_type": device_type,
|
| 179 |
|
| 180 |
"model_name": selected_model,
|
| 181 |
+
"context": context,
|
| 182 |
|
| 183 |
"score": score,
|
| 184 |
+
"context_score": context_score,
|
| 185 |
"timestamp": datetime.now(UTC).isoformat()
|
| 186 |
}
|
| 187 |
|
|
|
|
| 189 |
short_id = str(uuid.uuid4()) [:8]
|
| 190 |
|
| 191 |
file_name = (
|
| 192 |
+
f"{selected_model}_{user_id}_{timestamp}_{short_id}"
|
| 193 |
)
|
| 194 |
|
| 195 |
+
local_file = f"{file_name}.json"
|
| 196 |
|
| 197 |
with open(local_file, "w", encoding="utf-8") as f:
|
| 198 |
json.dump(annotation, f, indent=2)
|
| 199 |
|
| 200 |
api.upload_file(
|
| 201 |
path_or_fileobj=local_file,
|
| 202 |
+
path_in_repo=f"annotations/{local_file}",
|
| 203 |
repo_id=HF_DATASET_NAME,
|
| 204 |
repo_type="dataset"
|
| 205 |
)
|
|
|
|
| 215 |
tts_experience,
|
| 216 |
device_type,
|
| 217 |
selected_model,
|
| 218 |
+
context,
|
| 219 |
+
score,
|
| 220 |
+
context_score
|
| 221 |
):
|
| 222 |
+
try:
|
| 223 |
+
|
| 224 |
submit_annotation_hf(
|
| 225 |
user_id,
|
| 226 |
age_group,
|
|
|
|
| 229 |
tts_experience,
|
| 230 |
device_type,
|
| 231 |
selected_model,
|
| 232 |
+
context,
|
| 233 |
+
score,
|
| 234 |
+
context_score
|
| 235 |
)
|
| 236 |
|
| 237 |
+
return f"Annotation saved for {context}."
|
| 238 |
+
|
| 239 |
+
except Exception as e:
|
| 240 |
+
print("Submit error")
|
| 241 |
+
print(type(e).__name__)
|
| 242 |
+
print(e)
|
| 243 |
+
|
| 244 |
+
return f"ERROR: {e}"
|
| 245 |
+
|
| 246 |
+
def update_sentence(context):
|
| 247 |
+
|
| 248 |
+
return CONTEXTS[context]
|
| 249 |
+
|
| 250 |
|
| 251 |
# ============================
|
| 252 |
# SUBMIT BUTTON CONTROL
|
|
|
|
| 308 |
with gr.Row():
|
| 309 |
age_group = gr.Dropdown(
|
| 310 |
choices=[
|
| 311 |
+
"10-19",
|
| 312 |
+
"20-29",
|
| 313 |
+
"30-39",
|
| 314 |
+
"40-49",
|
| 315 |
+
"50-59",
|
| 316 |
+
"60-69",
|
| 317 |
"70+"
|
| 318 |
],
|
| 319 |
label="Age Group"
|
|
|
|
| 345 |
)
|
| 346 |
|
| 347 |
with gr.Row():
|
| 348 |
+
context_selector = gr.Dropdown(
|
| 349 |
+
choices=list(CONTEXTS.keys()),
|
| 350 |
+
value="Train station announcement",
|
| 351 |
+
label="Context"
|
| 352 |
+
)
|
| 353 |
+
|
| 354 |
+
sentence_display = gr.Textbox(
|
| 355 |
+
label = "Sentence",
|
| 356 |
+
interactive = False
|
| 357 |
)
|
| 358 |
|
| 359 |
model_selector = gr.Dropdown(
|
|
|
|
| 362 |
label = "Model"
|
| 363 |
)
|
| 364 |
|
| 365 |
+
context_selector.change(
|
| 366 |
+
update_sentence,
|
| 367 |
+
inputs=context_selector,
|
| 368 |
+
outputs=sentence_display
|
| 369 |
+
)
|
| 370 |
+
|
| 371 |
generate_btn = gr.Button("Generate Audio")
|
| 372 |
|
| 373 |
audio_player = gr.Audio(label="Generated audio")
|
|
|
|
| 376 |
generate_audio,
|
| 377 |
inputs=[
|
| 378 |
model_selector,
|
| 379 |
+
context_selector
|
| 380 |
],
|
| 381 |
outputs=[audio_player, audio_played]
|
| 382 |
)
|
| 383 |
|
| 384 |
with gr.Row():
|
| 385 |
score = gr.Dropdown(choices=["None","1", "2", "3", "4", "5"], value="None", label="MOS Score (1–5)")
|
| 386 |
+
score_context = gr.Dropdown(choices=["None","1", "2", "3", "4", "5"], value="None", label="Voice-context adequacy")
|
| 387 |
submit_btn = gr.Button("Submit Score", interactive=False)
|
| 388 |
|
| 389 |
status = gr.Textbox(label="Status", interactive=False)
|
|
|
|
| 423 |
native_language,
|
| 424 |
tts_experience,
|
| 425 |
device_type,
|
| 426 |
+
model_selector,
|
| 427 |
+
context_selector,
|
| 428 |
score
|
| 429 |
],
|
| 430 |
outputs=status
|