Spaces:
Running
Running
zach
commited on
Commit
Β·
048c3fc
1
Parent(s):
aad8b40
Move env var validation util to config, refactor to fix linting errors across project
Browse files- README.md +1 -0
- pyproject.toml +14 -1
- src/app.py +28 -26
- src/config.py +38 -11
- src/constants.py +18 -11
- src/custom_types.py +1 -1
- src/integrations/anthropic_api.py +28 -19
- src/integrations/elevenlabs_api.py +21 -15
- src/integrations/hume_api.py +27 -19
- src/utils.py +86 -103
README.md
CHANGED
@@ -39,6 +39,7 @@ Expressive TTS Arena/
|
|
39 |
β βββ app.py # Entry file
|
40 |
β βββ config.py # Global config and logger setup
|
41 |
β βββ constants.py # Global constants
|
|
|
42 |
β βββ theme.py # Custom Gradio Theme
|
43 |
β βββ utils.py # Utility functions
|
44 |
βββ .env.example
|
|
|
39 |
β βββ app.py # Entry file
|
40 |
β βββ config.py # Global config and logger setup
|
41 |
β βββ constants.py # Global constants
|
42 |
+
β βββ custom_types.py # Global custom types
|
43 |
β βββ theme.py # Custom Gradio Theme
|
44 |
β βββ utils.py # Utility functions
|
45 |
βββ .env.example
|
pyproject.toml
CHANGED
@@ -29,7 +29,20 @@ line-length = 120
|
|
29 |
src = ["src"]
|
30 |
|
31 |
[tool.ruff.lint]
|
32 |
-
ignore = [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
select = [
|
34 |
"A",
|
35 |
"ARG",
|
|
|
29 |
src = ["src"]
|
30 |
|
31 |
[tool.ruff.lint]
|
32 |
+
ignore = [
|
33 |
+
"B904",
|
34 |
+
"BLE001",
|
35 |
+
"D100",
|
36 |
+
"D104",
|
37 |
+
"EM101",
|
38 |
+
"EM102",
|
39 |
+
"FIX002",
|
40 |
+
"G004",
|
41 |
+
"PLR0913",
|
42 |
+
"PLR2004",
|
43 |
+
"TD002",
|
44 |
+
"TD003",
|
45 |
+
]
|
46 |
select = [
|
47 |
"A",
|
48 |
"ARG",
|
src/app.py
CHANGED
@@ -9,26 +9,26 @@ Users can compare the outputs and vote for their favorite in an interactive UI.
|
|
9 |
"""
|
10 |
|
11 |
# Standard Library Imports
|
12 |
-
from concurrent.futures import ThreadPoolExecutor
|
13 |
import time
|
|
|
14 |
from typing import Tuple, Union
|
15 |
|
16 |
# Third-Party Library Imports
|
17 |
import gradio as gr
|
18 |
|
19 |
# Local Application Imports
|
20 |
-
from src.config import AUDIO_DIR, logger
|
21 |
from src import constants
|
|
|
|
|
22 |
from src.integrations import (
|
23 |
AnthropicError,
|
24 |
ElevenLabsError,
|
25 |
-
generate_text_with_claude,
|
26 |
HumeError,
|
|
|
27 |
text_to_speech_with_elevenlabs,
|
28 |
text_to_speech_with_hume,
|
29 |
)
|
30 |
from src.theme import CustomTheme
|
31 |
-
from src.types import ComparisonType, OptionMap
|
32 |
from src.utils import (
|
33 |
choose_providers,
|
34 |
create_shuffled_tts_options,
|
@@ -66,7 +66,7 @@ def generate_text(
|
|
66 |
logger.info(f"Generated text ({len(generated_text)} characters).")
|
67 |
return gr.update(value=generated_text), generated_text
|
68 |
except AnthropicError as ae:
|
69 |
-
logger.error(f"AnthropicError while generating text: {
|
70 |
raise gr.Error(
|
71 |
f'There was an issue communicating with the Anthropic API: "{ae.message}"'
|
72 |
)
|
@@ -94,7 +94,8 @@ def synthesize_speech(
|
|
94 |
Args:
|
95 |
character_description (str): The description of the character used for generating the voice.
|
96 |
text (str): The text content to be synthesized into speech.
|
97 |
-
generated_text_state (str): The previously generated text state, used to determine if the text has
|
|
|
98 |
|
99 |
Returns:
|
100 |
Tuple containing:
|
@@ -118,7 +119,7 @@ def synthesize_speech(
|
|
118 |
|
119 |
# Select 2 TTS providers based on whether the text has been modified.
|
120 |
text_modified = text != generated_text_state
|
121 |
-
|
122 |
text_modified, character_description
|
123 |
)
|
124 |
|
@@ -151,9 +152,9 @@ def synthesize_speech(
|
|
151 |
generation_id_b, audio_b = future_audio_b.result()
|
152 |
|
153 |
# Shuffle options so that placement of options in the UI will always be random
|
154 |
-
|
155 |
-
|
156 |
-
)
|
157 |
|
158 |
option_a_audio = options_map["option_a"]["audio_file_path"]
|
159 |
option_b_audio = options_map["option_b"]["audio_file_path"]
|
@@ -162,18 +163,17 @@ def synthesize_speech(
|
|
162 |
gr.update(value=option_a_audio, visible=True, autoplay=True),
|
163 |
gr.update(value=option_b_audio, visible=True),
|
164 |
options_map,
|
165 |
-
comparison_type,
|
166 |
text_modified,
|
167 |
text,
|
168 |
character_description,
|
169 |
)
|
170 |
except ElevenLabsError as ee:
|
171 |
-
logger.error(f"ElevenLabsError while synthesizing speech from text: {
|
172 |
raise gr.Error(
|
173 |
f'There was an issue communicating with the Elevenlabs API: "{ee.message}"'
|
174 |
)
|
175 |
except HumeError as he:
|
176 |
-
logger.error(f"HumeError while synthesizing speech from text: {
|
177 |
raise gr.Error(
|
178 |
f'There was an issue communicating with the Hume API: "{he.message}"'
|
179 |
)
|
@@ -186,7 +186,6 @@ def vote(
|
|
186 |
vote_submitted: bool,
|
187 |
option_map: OptionMap,
|
188 |
clicked_option_button: str,
|
189 |
-
comparison_type: ComparisonType,
|
190 |
text_modified: bool,
|
191 |
character_description: str,
|
192 |
text: str,
|
@@ -222,7 +221,6 @@ def vote(
|
|
222 |
submit_voting_results(
|
223 |
option_map,
|
224 |
selected_option,
|
225 |
-
comparison_type,
|
226 |
text_modified,
|
227 |
character_description,
|
228 |
text,
|
@@ -272,7 +270,10 @@ def reset_ui() -> Tuple[gr.update, gr.update, gr.update, gr.update, None, bool]:
|
|
272 |
|
273 |
|
274 |
def build_input_section() -> Tuple[gr.Dropdown, gr.Textbox, gr.Button]:
|
275 |
-
"""
|
|
|
|
|
|
|
276 |
sample_character_description_dropdown = gr.Dropdown(
|
277 |
choices=list(constants.SAMPLE_CHARACTER_DESCRIPTIONS.keys()),
|
278 |
label="Choose a sample character description",
|
@@ -298,7 +299,9 @@ def build_input_section() -> Tuple[gr.Dropdown, gr.Textbox, gr.Button]:
|
|
298 |
def build_output_section() -> (
|
299 |
Tuple[gr.Textbox, gr.Button, gr.Audio, gr.Audio, gr.Button, gr.Button]
|
300 |
):
|
301 |
-
"""
|
|
|
|
|
302 |
text_input = gr.Textbox(
|
303 |
label="Input Text",
|
304 |
placeholder="Enter or generate text for synthesis...",
|
@@ -348,11 +351,15 @@ def build_gradio_interface() -> gr.Blocks:
|
|
348 |
gr.Markdown("# Expressive TTS Arena")
|
349 |
gr.Markdown(
|
350 |
"""
|
351 |
-
1. **Choose or enter a character description**: Select a sample from the list or enter your own to guide
|
352 |
-
|
353 |
-
|
|
|
|
|
|
|
354 |
4. **Listen & compare**: Play both audio options and assess their expressiveness.
|
355 |
-
5. **Vote for the best**: Click **"Select Option A"** or **"Select Option B"** to choose the most
|
|
|
356 |
"""
|
357 |
)
|
358 |
|
@@ -384,8 +391,6 @@ def build_gradio_interface() -> gr.Blocks:
|
|
384 |
# Track whether text that was used was generated or modified/custom
|
385 |
text_modified_state = gr.State()
|
386 |
|
387 |
-
# Track comparison type (which set of providers are being compared)
|
388 |
-
comparison_type_state = gr.State()
|
389 |
# Track option map (option A and option B are randomized)
|
390 |
option_map_state = gr.State()
|
391 |
|
@@ -450,7 +455,6 @@ def build_gradio_interface() -> gr.Blocks:
|
|
450 |
option_a_audio_player,
|
451 |
option_b_audio_player,
|
452 |
option_map_state,
|
453 |
-
comparison_type_state,
|
454 |
text_modified_state,
|
455 |
text_state,
|
456 |
character_description_state,
|
@@ -472,7 +476,6 @@ def build_gradio_interface() -> gr.Blocks:
|
|
472 |
vote_submitted_state,
|
473 |
option_map_state,
|
474 |
vote_button_a,
|
475 |
-
comparison_type_state,
|
476 |
text_modified_state,
|
477 |
character_description_state,
|
478 |
text_state,
|
@@ -490,7 +493,6 @@ def build_gradio_interface() -> gr.Blocks:
|
|
490 |
vote_submitted_state,
|
491 |
option_map_state,
|
492 |
vote_button_b,
|
493 |
-
comparison_type_state,
|
494 |
text_modified_state,
|
495 |
character_description_state,
|
496 |
text_state,
|
|
|
9 |
"""
|
10 |
|
11 |
# Standard Library Imports
|
|
|
12 |
import time
|
13 |
+
from concurrent.futures import ThreadPoolExecutor
|
14 |
from typing import Tuple, Union
|
15 |
|
16 |
# Third-Party Library Imports
|
17 |
import gradio as gr
|
18 |
|
19 |
# Local Application Imports
|
|
|
20 |
from src import constants
|
21 |
+
from src.config import AUDIO_DIR, logger
|
22 |
+
from src.custom_types import ComparisonType, Option, OptionMap
|
23 |
from src.integrations import (
|
24 |
AnthropicError,
|
25 |
ElevenLabsError,
|
|
|
26 |
HumeError,
|
27 |
+
generate_text_with_claude,
|
28 |
text_to_speech_with_elevenlabs,
|
29 |
text_to_speech_with_hume,
|
30 |
)
|
31 |
from src.theme import CustomTheme
|
|
|
32 |
from src.utils import (
|
33 |
choose_providers,
|
34 |
create_shuffled_tts_options,
|
|
|
66 |
logger.info(f"Generated text ({len(generated_text)} characters).")
|
67 |
return gr.update(value=generated_text), generated_text
|
68 |
except AnthropicError as ae:
|
69 |
+
logger.error(f"AnthropicError while generating text: {ae!s}")
|
70 |
raise gr.Error(
|
71 |
f'There was an issue communicating with the Anthropic API: "{ae.message}"'
|
72 |
)
|
|
|
94 |
Args:
|
95 |
character_description (str): The description of the character used for generating the voice.
|
96 |
text (str): The text content to be synthesized into speech.
|
97 |
+
generated_text_state (str): The previously generated text state, used to determine if the text has
|
98 |
+
been modified.
|
99 |
|
100 |
Returns:
|
101 |
Tuple containing:
|
|
|
119 |
|
120 |
# Select 2 TTS providers based on whether the text has been modified.
|
121 |
text_modified = text != generated_text_state
|
122 |
+
provider_a, provider_b = choose_providers(
|
123 |
text_modified, character_description
|
124 |
)
|
125 |
|
|
|
152 |
generation_id_b, audio_b = future_audio_b.result()
|
153 |
|
154 |
# Shuffle options so that placement of options in the UI will always be random
|
155 |
+
option_a = Option(provider=provider_a, audio=audio_a, generation_id=generation_id_a)
|
156 |
+
option_b = Option(provider=provider_b, audio=audio_b, generation_id=generation_id_b)
|
157 |
+
options_map: OptionMap = create_shuffled_tts_options(option_a, option_b)
|
158 |
|
159 |
option_a_audio = options_map["option_a"]["audio_file_path"]
|
160 |
option_b_audio = options_map["option_b"]["audio_file_path"]
|
|
|
163 |
gr.update(value=option_a_audio, visible=True, autoplay=True),
|
164 |
gr.update(value=option_b_audio, visible=True),
|
165 |
options_map,
|
|
|
166 |
text_modified,
|
167 |
text,
|
168 |
character_description,
|
169 |
)
|
170 |
except ElevenLabsError as ee:
|
171 |
+
logger.error(f"ElevenLabsError while synthesizing speech from text: {ee!s}")
|
172 |
raise gr.Error(
|
173 |
f'There was an issue communicating with the Elevenlabs API: "{ee.message}"'
|
174 |
)
|
175 |
except HumeError as he:
|
176 |
+
logger.error(f"HumeError while synthesizing speech from text: {he!s}")
|
177 |
raise gr.Error(
|
178 |
f'There was an issue communicating with the Hume API: "{he.message}"'
|
179 |
)
|
|
|
186 |
vote_submitted: bool,
|
187 |
option_map: OptionMap,
|
188 |
clicked_option_button: str,
|
|
|
189 |
text_modified: bool,
|
190 |
character_description: str,
|
191 |
text: str,
|
|
|
221 |
submit_voting_results(
|
222 |
option_map,
|
223 |
selected_option,
|
|
|
224 |
text_modified,
|
225 |
character_description,
|
226 |
text,
|
|
|
270 |
|
271 |
|
272 |
def build_input_section() -> Tuple[gr.Dropdown, gr.Textbox, gr.Button]:
|
273 |
+
"""
|
274 |
+
Builds the input section including the sample character description dropdown, character
|
275 |
+
description input, and generate text button.
|
276 |
+
"""
|
277 |
sample_character_description_dropdown = gr.Dropdown(
|
278 |
choices=list(constants.SAMPLE_CHARACTER_DESCRIPTIONS.keys()),
|
279 |
label="Choose a sample character description",
|
|
|
299 |
def build_output_section() -> (
|
300 |
Tuple[gr.Textbox, gr.Button, gr.Audio, gr.Audio, gr.Button, gr.Button]
|
301 |
):
|
302 |
+
"""
|
303 |
+
Builds the output section including text input, audio players, and vote buttons.
|
304 |
+
"""
|
305 |
text_input = gr.Textbox(
|
306 |
label="Input Text",
|
307 |
placeholder="Enter or generate text for synthesis...",
|
|
|
351 |
gr.Markdown("# Expressive TTS Arena")
|
352 |
gr.Markdown(
|
353 |
"""
|
354 |
+
1. **Choose or enter a character description**: Select a sample from the list or enter your own to guide
|
355 |
+
text and voice generation.
|
356 |
+
2. **Generate text**: Click **"Generate Text"** to create dialogue based on the character. The generated
|
357 |
+
text will appear in the input field automaticallyβedit it if needed.
|
358 |
+
3. **Synthesize speech**: Click **"Synthesize Speech"** to send your text and character description to two
|
359 |
+
TTS APIs. Each API generates a voice and synthesizes speech in that voice.
|
360 |
4. **Listen & compare**: Play both audio options and assess their expressiveness.
|
361 |
+
5. **Vote for the best**: Click **"Select Option A"** or **"Select Option B"** to choose the most
|
362 |
+
expressive output.
|
363 |
"""
|
364 |
)
|
365 |
|
|
|
391 |
# Track whether text that was used was generated or modified/custom
|
392 |
text_modified_state = gr.State()
|
393 |
|
|
|
|
|
394 |
# Track option map (option A and option B are randomized)
|
395 |
option_map_state = gr.State()
|
396 |
|
|
|
455 |
option_a_audio_player,
|
456 |
option_b_audio_player,
|
457 |
option_map_state,
|
|
|
458 |
text_modified_state,
|
459 |
text_state,
|
460 |
character_description_state,
|
|
|
476 |
vote_submitted_state,
|
477 |
option_map_state,
|
478 |
vote_button_a,
|
|
|
479 |
text_modified_state,
|
480 |
character_description_state,
|
481 |
text_state,
|
|
|
493 |
vote_submitted_state,
|
494 |
option_map_state,
|
495 |
vote_button_b,
|
|
|
496 |
text_modified_state,
|
497 |
character_description_state,
|
498 |
text_state,
|
src/config.py
CHANGED
@@ -13,25 +13,20 @@ Key Features:
|
|
13 |
# Standard Library Imports
|
14 |
import logging
|
15 |
import os
|
|
|
16 |
|
17 |
# Third-Party Library Imports
|
18 |
from dotenv import load_dotenv
|
19 |
|
20 |
-
|
21 |
# Determine the environment (defaults to "dev" if not explicitly set)
|
22 |
APP_ENV = os.getenv("APP_ENV", "dev").lower()
|
23 |
if APP_ENV not in {"dev", "prod"}:
|
24 |
-
print(f'Warning: Invalid APP_ENV "{APP_ENV}". Defaulting to "dev".')
|
25 |
APP_ENV = "dev"
|
26 |
|
27 |
|
28 |
# In development, load environment variables from .env file (not used in production)
|
29 |
-
if APP_ENV == "dev":
|
30 |
-
|
31 |
-
# Load environment variables
|
32 |
-
load_dotenv(".env", override=True)
|
33 |
-
else:
|
34 |
-
print("Warning: .env file not found. Using system environment variables.")
|
35 |
|
36 |
|
37 |
# Enable debug mode if in development (or if explicitly set in env variables)
|
@@ -47,10 +42,42 @@ logger.info(f'App running in "{APP_ENV}" mode.')
|
|
47 |
logger.info(f'Debug mode is {"enabled" if DEBUG else "disabled"}.')
|
48 |
|
49 |
if DEBUG:
|
50 |
-
logger.debug(
|
51 |
|
52 |
|
53 |
# Define the directory for audio files relative to the project root
|
54 |
-
AUDIO_DIR =
|
55 |
-
|
56 |
logger.info(f"Audio directory set to {AUDIO_DIR}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
# Standard Library Imports
|
14 |
import logging
|
15 |
import os
|
16 |
+
from pathlib import Path
|
17 |
|
18 |
# Third-Party Library Imports
|
19 |
from dotenv import load_dotenv
|
20 |
|
|
|
21 |
# Determine the environment (defaults to "dev" if not explicitly set)
|
22 |
APP_ENV = os.getenv("APP_ENV", "dev").lower()
|
23 |
if APP_ENV not in {"dev", "prod"}:
|
|
|
24 |
APP_ENV = "dev"
|
25 |
|
26 |
|
27 |
# In development, load environment variables from .env file (not used in production)
|
28 |
+
if APP_ENV == "dev" and Path(".env").exists():
|
29 |
+
load_dotenv(".env", override=True)
|
|
|
|
|
|
|
|
|
30 |
|
31 |
|
32 |
# Enable debug mode if in development (or if explicitly set in env variables)
|
|
|
42 |
logger.info(f'Debug mode is {"enabled" if DEBUG else "disabled"}.')
|
43 |
|
44 |
if DEBUG:
|
45 |
+
logger.debug("DEBUG mode enabled.")
|
46 |
|
47 |
|
48 |
# Define the directory for audio files relative to the project root
|
49 |
+
AUDIO_DIR = Path.cwd() / "static" / "audio"
|
50 |
+
AUDIO_DIR.mkdir(parents=True, exist_ok=True)
|
51 |
logger.info(f"Audio directory set to {AUDIO_DIR}")
|
52 |
+
|
53 |
+
|
54 |
+
def validate_env_var(var_name: str) -> str:
|
55 |
+
"""
|
56 |
+
Validates that an environment variable is set and returns its value.
|
57 |
+
|
58 |
+
Args:
|
59 |
+
var_name (str): The name of the environment variable to validate.
|
60 |
+
|
61 |
+
Returns:
|
62 |
+
str: The value of the environment variable.
|
63 |
+
|
64 |
+
Raises:
|
65 |
+
ValueError: If the environment variable is not set.
|
66 |
+
|
67 |
+
Examples:
|
68 |
+
>>> import os
|
69 |
+
>>> os.environ["EXAMPLE_VAR"] = "example_value"
|
70 |
+
>>> validate_env_var("EXAMPLE_VAR")
|
71 |
+
'example_value'
|
72 |
+
|
73 |
+
>>> validate_env_var("MISSING_VAR")
|
74 |
+
Traceback (most recent call last):
|
75 |
+
...
|
76 |
+
ValueError: MISSING_VAR is not set. Please ensure it is defined in your environment variables.
|
77 |
+
"""
|
78 |
+
value = os.environ.get(var_name, "")
|
79 |
+
if not value:
|
80 |
+
raise ValueError(
|
81 |
+
f"{var_name} is not set. Please ensure it is defined in your environment variables."
|
82 |
+
)
|
83 |
+
return value
|
src/constants.py
CHANGED
@@ -8,8 +8,10 @@ This module defines global constants used throughout the project.
|
|
8 |
from typing import List
|
9 |
|
10 |
# Third-Party Library Imports
|
11 |
-
from src.
|
12 |
|
|
|
|
|
13 |
|
14 |
# UI constants
|
15 |
HUME_AI: TTSProviderName = "Hume AI"
|
@@ -35,23 +37,28 @@ SELECT_OPTION_B: str = "Select Option B"
|
|
35 |
# inspiration for generating creative text for expressive TTS, and generating novel voices.
|
36 |
SAMPLE_CHARACTER_DESCRIPTIONS: dict = {
|
37 |
"π Stranded Astronaut": (
|
38 |
-
"A lone astronaut whose voice mirrors the silent vastness of spaceβa low, steady tone imbued
|
39 |
-
"It carries the measured resolve of someone sending a final
|
|
|
40 |
),
|
41 |
"π Timeless Poet": (
|
42 |
-
"An ageless poet with a voice that flows like gentle verseβa soft, reflective tone marked by
|
43 |
-
"It speaks with the measured cadence of classic sonnets, evoking both the
|
|
|
44 |
),
|
45 |
"π± Whimsical Feline": (
|
46 |
-
"A mischievous cat whose voice is playful yet mysteriousβlight, quick-witted, and infused with
|
47 |
-
"It hints at secret adventures and hidden charm, balancing exuberance with
|
|
|
48 |
),
|
49 |
"π₯ Revolutionary Orator": (
|
50 |
-
"A defiant orator whose voice builds from quiet determination to passionate fervorβa clear,
|
51 |
-
"It starts measured and resolute, then rises
|
|
|
52 |
),
|
53 |
"π» Haunted Keeper": (
|
54 |
-
"A solitary lighthouse keeper with a voice that carries the weight of forgotten stormsβa soft,
|
55 |
-
"It speaks as if whispering long-held secrets in the dark,
|
|
|
56 |
),
|
57 |
}
|
|
|
8 |
from typing import List
|
9 |
|
10 |
# Third-Party Library Imports
|
11 |
+
from src.custom_types import ComparisonType, OptionKey, OptionLabel, TTSProviderName
|
12 |
|
13 |
+
CLIENT_ERROR_CODE = 400
|
14 |
+
SERVER_ERROR_CODE = 500
|
15 |
|
16 |
# UI constants
|
17 |
HUME_AI: TTSProviderName = "Hume AI"
|
|
|
37 |
# inspiration for generating creative text for expressive TTS, and generating novel voices.
|
38 |
SAMPLE_CHARACTER_DESCRIPTIONS: dict = {
|
39 |
"π Stranded Astronaut": (
|
40 |
+
"A lone astronaut whose voice mirrors the silent vastness of spaceβa low, steady tone imbued "
|
41 |
+
"with isolation and quiet wonder. It carries the measured resolve of someone sending a final "
|
42 |
+
"transmission, with an undercurrent of wistful melancholy."
|
43 |
),
|
44 |
"π Timeless Poet": (
|
45 |
+
"An ageless poet with a voice that flows like gentle verseβa soft, reflective tone marked by "
|
46 |
+
"deliberate pauses. It speaks with the measured cadence of classic sonnets, evoking both the "
|
47 |
+
"fragile beauty of time and heartfelt introspection."
|
48 |
),
|
49 |
"π± Whimsical Feline": (
|
50 |
+
"A mischievous cat whose voice is playful yet mysteriousβlight, quick-witted, and infused with "
|
51 |
+
"an enchanting purr. It hints at secret adventures and hidden charm, balancing exuberance with "
|
52 |
+
"a subtle, smooth allure."
|
53 |
),
|
54 |
"π₯ Revolutionary Orator": (
|
55 |
+
"A defiant orator whose voice builds from quiet determination to passionate fervorβa clear, "
|
56 |
+
"commanding tone that resonates with conviction. It starts measured and resolute, then rises "
|
57 |
+
"to a crescendo of fervor, punctuated by deliberate pauses that emphasize each rallying cry."
|
58 |
),
|
59 |
"π» Haunted Keeper": (
|
60 |
+
"A solitary lighthouse keeper with a voice that carries the weight of forgotten stormsβa soft, "
|
61 |
+
"measured tone with an echo of sorrow. It speaks as if whispering long-held secrets in the dark, "
|
62 |
+
"blending quiet melancholy with an air of enduring mystery."
|
63 |
),
|
64 |
}
|
src/custom_types.py
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
"""
|
2 |
-
|
3 |
|
4 |
This module defines custom types for the application.
|
5 |
"""
|
|
|
1 |
"""
|
2 |
+
custom_types.py
|
3 |
|
4 |
This module defines custom types for the application.
|
5 |
"""
|
src/integrations/anthropic_api.py
CHANGED
@@ -19,25 +19,26 @@ Functions:
|
|
19 |
"""
|
20 |
|
21 |
# Standard Library Imports
|
22 |
-
from dataclasses import dataclass
|
23 |
import logging
|
|
|
24 |
from typing import List, Optional, Union
|
25 |
|
26 |
# Third-Party Library Imports
|
27 |
-
from anthropic import
|
28 |
from anthropic.types import Message, ModelParam, TextBlock
|
29 |
-
from tenacity import
|
30 |
|
31 |
# Local Application Imports
|
32 |
-
from src.config import logger
|
33 |
-
from src.
|
|
|
34 |
|
35 |
|
36 |
@dataclass(frozen=True)
|
37 |
class AnthropicConfig:
|
38 |
"""Immutable configuration for interacting with the Anthropic API."""
|
39 |
|
40 |
-
api_key: str =
|
41 |
model: ModelParam = "claude-3-5-sonnet-latest"
|
42 |
max_tokens: int = 150
|
43 |
system_prompt: Optional[str] = (
|
@@ -47,13 +48,16 @@ class AnthropicConfig:
|
|
47 |
def __post_init__(self):
|
48 |
# Validate that required attributes are set
|
49 |
if not self.api_key:
|
50 |
-
|
|
|
51 |
if not self.model:
|
52 |
raise ValueError("Anthropic Model is not set.")
|
53 |
if not self.max_tokens:
|
54 |
raise ValueError("Anthropic Max Tokens is not set.")
|
55 |
if self.system_prompt is None:
|
56 |
-
system_prompt: str = f"""You are an expert at generating micro-content optimized for text-to-speech
|
|
|
|
|
57 |
CRITICAL LENGTH CONSTRAINTS:
|
58 |
|
59 |
Maximum length: {self.max_tokens} tokens (approximately 400 characters)
|
@@ -84,7 +88,8 @@ Opening hook (50-75 characters)
|
|
84 |
Emotional journey (200-250 characters)
|
85 |
Resolution (75-100 characters)
|
86 |
|
87 |
-
MANDATORY: If you find yourself reaching 300 characters, immediately begin your conclusion regardless of where you
|
|
|
88 |
Remember: A shorter, complete response is ALWAYS better than a longer, truncated one."""
|
89 |
object.__setattr__(self, "system_prompt", system_prompt)
|
90 |
|
@@ -110,12 +115,13 @@ Remember: A shorter, complete response is ALWAYS better than a longer, truncated
|
|
110 |
Returns:
|
111 |
str: The prompt to be passed to the Anthropic API.
|
112 |
"""
|
113 |
-
|
114 |
f"Character Description: {character_description}\n\n"
|
115 |
-
"Based on the above character description, please generate a line of dialogue that captures the
|
116 |
-
"
|
|
|
|
|
117 |
)
|
118 |
-
return prompt
|
119 |
|
120 |
|
121 |
class AnthropicError(Exception):
|
@@ -198,12 +204,15 @@ def generate_text_with_claude(character_description: str) -> str:
|
|
198 |
return str(blocks or "No content generated.")
|
199 |
|
200 |
except Exception as e:
|
201 |
-
if
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
|
|
|
|
|
|
207 |
raise AnthropicError(
|
208 |
message=(f"{e.message}"),
|
209 |
original_exception=e,
|
|
|
19 |
"""
|
20 |
|
21 |
# Standard Library Imports
|
|
|
22 |
import logging
|
23 |
+
from dataclasses import dataclass
|
24 |
from typing import List, Optional, Union
|
25 |
|
26 |
# Third-Party Library Imports
|
27 |
+
from anthropic import Anthropic, APIError
|
28 |
from anthropic.types import Message, ModelParam, TextBlock
|
29 |
+
from tenacity import after_log, before_log, retry, stop_after_attempt, wait_fixed
|
30 |
|
31 |
# Local Application Imports
|
32 |
+
from src.config import logger, validate_env_var
|
33 |
+
from src.constants import CLIENT_ERROR_CODE, SERVER_ERROR_CODE
|
34 |
+
from src.utils import truncate_text
|
35 |
|
36 |
|
37 |
@dataclass(frozen=True)
|
38 |
class AnthropicConfig:
|
39 |
"""Immutable configuration for interacting with the Anthropic API."""
|
40 |
|
41 |
+
api_key: Optional[str] = None
|
42 |
model: ModelParam = "claude-3-5-sonnet-latest"
|
43 |
max_tokens: int = 150
|
44 |
system_prompt: Optional[str] = (
|
|
|
48 |
def __post_init__(self):
|
49 |
# Validate that required attributes are set
|
50 |
if not self.api_key:
|
51 |
+
api_key = validate_env_var("ANTHROPIC_API_KEY")
|
52 |
+
object.__setattr__(self, "api_key", api_key)
|
53 |
if not self.model:
|
54 |
raise ValueError("Anthropic Model is not set.")
|
55 |
if not self.max_tokens:
|
56 |
raise ValueError("Anthropic Max Tokens is not set.")
|
57 |
if self.system_prompt is None:
|
58 |
+
system_prompt: str = f"""You are an expert at generating micro-content optimized for text-to-speech
|
59 |
+
synthesis. Your absolute priority is delivering complete, untruncated responses within strict length limits.
|
60 |
+
|
61 |
CRITICAL LENGTH CONSTRAINTS:
|
62 |
|
63 |
Maximum length: {self.max_tokens} tokens (approximately 400 characters)
|
|
|
88 |
Emotional journey (200-250 characters)
|
89 |
Resolution (75-100 characters)
|
90 |
|
91 |
+
MANDATORY: If you find yourself reaching 300 characters, immediately begin your conclusion regardless of where you
|
92 |
+
are in the narrative.
|
93 |
Remember: A shorter, complete response is ALWAYS better than a longer, truncated one."""
|
94 |
object.__setattr__(self, "system_prompt", system_prompt)
|
95 |
|
|
|
115 |
Returns:
|
116 |
str: The prompt to be passed to the Anthropic API.
|
117 |
"""
|
118 |
+
return (
|
119 |
f"Character Description: {character_description}\n\n"
|
120 |
+
"Based on the above character description, please generate a line of dialogue that captures the "
|
121 |
+
"character's unique personality, emotional depth, and distinctive tone. The response should sound "
|
122 |
+
"like something the character would naturally say, reflecting their background and emotional state, "
|
123 |
+
"and be fully developed for text-to-speech synthesis."
|
124 |
)
|
|
|
125 |
|
126 |
|
127 |
class AnthropicError(Exception):
|
|
|
204 |
return str(blocks or "No content generated.")
|
205 |
|
206 |
except Exception as e:
|
207 |
+
if (
|
208 |
+
isinstance(e, APIError)
|
209 |
+
and e.status_code >= CLIENT_ERROR_CODE and e.status_code < SERVER_ERROR_CODE
|
210 |
+
):
|
211 |
+
raise UnretryableAnthropicError(
|
212 |
+
message=f"\"{e.body['error']['message']}\"",
|
213 |
+
original_exception=e,
|
214 |
+
) from e
|
215 |
+
|
216 |
raise AnthropicError(
|
217 |
message=(f"{e.message}"),
|
218 |
original_exception=e,
|
src/integrations/elevenlabs_api.py
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
"""
|
2 |
elevenlabs_api.py
|
3 |
|
4 |
-
This file defines the interaction with the ElevenLabs text-to-speech (TTS) API using the
|
5 |
-
It includes functionality for API request handling and processing API responses.
|
6 |
|
7 |
Key Features:
|
8 |
- Encapsulates all logic related to the ElevenLabs TTS API.
|
@@ -20,32 +20,34 @@ Functions:
|
|
20 |
"""
|
21 |
|
22 |
# Standard Library Imports
|
23 |
-
from dataclasses import dataclass
|
24 |
import logging
|
25 |
import random
|
|
|
26 |
from typing import Optional, Tuple
|
27 |
|
28 |
# Third-Party Library Imports
|
29 |
from elevenlabs import ElevenLabs, TextToVoiceCreatePreviewsRequestOutputFormat
|
30 |
from elevenlabs.core import ApiError
|
31 |
-
from tenacity import
|
32 |
|
33 |
# Local Application Imports
|
34 |
-
from src.config import logger
|
35 |
-
from src.
|
|
|
36 |
|
37 |
|
38 |
@dataclass(frozen=True)
|
39 |
class ElevenLabsConfig:
|
40 |
"""Immutable configuration for interacting with the ElevenLabs TTS API."""
|
41 |
|
42 |
-
api_key: str =
|
43 |
output_format: TextToVoiceCreatePreviewsRequestOutputFormat = "mp3_44100_128"
|
44 |
|
45 |
def __post_init__(self):
|
46 |
# Validate that required attributes are set
|
47 |
if not self.api_key:
|
48 |
-
|
|
|
49 |
|
50 |
@property
|
51 |
def client(self) -> ElevenLabs:
|
@@ -97,7 +99,8 @@ def text_to_speech_with_elevenlabs(
|
|
97 |
|
98 |
Returns:
|
99 |
Tuple[None, str]: A tuple containing:
|
100 |
-
- generation_id (None): We do not record the generation ID for ElevenLabs, but return None for uniformity
|
|
|
101 |
- file_path (str): The relative file path to the audio file where the synthesized speech was saved.
|
102 |
|
103 |
Raises:
|
@@ -132,12 +135,15 @@ def text_to_speech_with_elevenlabs(
|
|
132 |
return None, audio_file_path
|
133 |
|
134 |
except Exception as e:
|
135 |
-
if
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
|
|
|
|
|
|
141 |
raise ElevenLabsError(
|
142 |
message=f"{e}",
|
143 |
original_exception=e,
|
|
|
1 |
"""
|
2 |
elevenlabs_api.py
|
3 |
|
4 |
+
This file defines the interaction with the ElevenLabs text-to-speech (TTS) API using the
|
5 |
+
ElevenLabs Python SDK. It includes functionality for API request handling and processing API responses.
|
6 |
|
7 |
Key Features:
|
8 |
- Encapsulates all logic related to the ElevenLabs TTS API.
|
|
|
20 |
"""
|
21 |
|
22 |
# Standard Library Imports
|
|
|
23 |
import logging
|
24 |
import random
|
25 |
+
from dataclasses import dataclass
|
26 |
from typing import Optional, Tuple
|
27 |
|
28 |
# Third-Party Library Imports
|
29 |
from elevenlabs import ElevenLabs, TextToVoiceCreatePreviewsRequestOutputFormat
|
30 |
from elevenlabs.core import ApiError
|
31 |
+
from tenacity import after_log, before_log, retry, stop_after_attempt, wait_fixed
|
32 |
|
33 |
# Local Application Imports
|
34 |
+
from src.config import logger, validate_env_var
|
35 |
+
from src.constants import CLIENT_ERROR_CODE, SERVER_ERROR_CODE
|
36 |
+
from src.utils import save_base64_audio_to_file
|
37 |
|
38 |
|
39 |
@dataclass(frozen=True)
|
40 |
class ElevenLabsConfig:
|
41 |
"""Immutable configuration for interacting with the ElevenLabs TTS API."""
|
42 |
|
43 |
+
api_key: Optional[str] = None
|
44 |
output_format: TextToVoiceCreatePreviewsRequestOutputFormat = "mp3_44100_128"
|
45 |
|
46 |
def __post_init__(self):
|
47 |
# Validate that required attributes are set
|
48 |
if not self.api_key:
|
49 |
+
api_key = validate_env_var("ELEVENLABS_API_KEY")
|
50 |
+
object.__setattr__(self, "api_key", api_key)
|
51 |
|
52 |
@property
|
53 |
def client(self) -> ElevenLabs:
|
|
|
99 |
|
100 |
Returns:
|
101 |
Tuple[None, str]: A tuple containing:
|
102 |
+
- generation_id (None): We do not record the generation ID for ElevenLabs, but return None for uniformity
|
103 |
+
across TTS integrations
|
104 |
- file_path (str): The relative file path to the audio file where the synthesized speech was saved.
|
105 |
|
106 |
Raises:
|
|
|
135 |
return None, audio_file_path
|
136 |
|
137 |
except Exception as e:
|
138 |
+
if (
|
139 |
+
isinstance(e, ApiError)
|
140 |
+
and e.status_code >= CLIENT_ERROR_CODE and e.status_code < SERVER_ERROR_CODE
|
141 |
+
):
|
142 |
+
raise UnretryableElevenLabsError(
|
143 |
+
message=f"{e.body['detail']['message']}",
|
144 |
+
original_exception=e,
|
145 |
+
) from e
|
146 |
+
|
147 |
raise ElevenLabsError(
|
148 |
message=f"{e}",
|
149 |
original_exception=e,
|
src/integrations/hume_api.py
CHANGED
@@ -19,21 +19,19 @@ Functions:
|
|
19 |
"""
|
20 |
|
21 |
# Standard Library Imports
|
22 |
-
from dataclasses import dataclass
|
23 |
import logging
|
24 |
-
import
|
25 |
-
import random
|
26 |
from typing import Any, Dict, Literal, Optional, Tuple, Union
|
27 |
|
28 |
# Third-Party Library Imports
|
29 |
import requests
|
30 |
from requests.exceptions import HTTPError
|
31 |
-
from tenacity import
|
32 |
|
33 |
# Local Application Imports
|
34 |
-
from src.config import logger
|
35 |
-
from src.
|
36 |
-
|
37 |
|
38 |
HumeSupportedFileFormat = Literal["mp3", "pcm", "wav"]
|
39 |
""" Support audio file formats for the Hume TTS API"""
|
@@ -43,7 +41,7 @@ HumeSupportedFileFormat = Literal["mp3", "pcm", "wav"]
|
|
43 |
class HumeConfig:
|
44 |
"""Immutable configuration for interacting with the Hume TTS API."""
|
45 |
|
46 |
-
api_key: str =
|
47 |
url: str = "https://test-api.hume.ai/v0/tts/octave"
|
48 |
headers: dict = None
|
49 |
file_format: HumeSupportedFileFormat = "mp3"
|
@@ -51,7 +49,8 @@ class HumeConfig:
|
|
51 |
def __post_init__(self):
|
52 |
# Validate required attributes
|
53 |
if not self.api_key:
|
54 |
-
|
|
|
55 |
if not self.url:
|
56 |
raise ValueError("Hume TTS endpoint URL is not set.")
|
57 |
if not self.file_format:
|
@@ -118,17 +117,19 @@ def text_to_speech_with_hume(
|
|
118 |
|
119 |
Returns:
|
120 |
Union[Tuple[str, str], Tuple[str, str, str, str]]:
|
121 |
-
- If num_generations == 1:
|
122 |
-
- If num_generations == 2:
|
123 |
|
124 |
Raises:
|
125 |
ValueError: If num_generations is not 1 or 2.
|
126 |
HumeError: If there is an error communicating with the Hume TTS API or parsing its response.
|
127 |
UnretryableHumeError: If a client-side HTTP error (status code in the 4xx range) is encountered.
|
128 |
-
Exception: Any other exceptions raised during the request or processing will be wrapped and
|
|
|
129 |
"""
|
130 |
logger.debug(
|
131 |
-
f"Processing TTS with Hume. Prompt length: {len(character_description)} characters.
|
|
|
132 |
)
|
133 |
|
134 |
if num_generations < 1 or num_generations > 2:
|
@@ -170,12 +171,19 @@ def text_to_speech_with_hume(
|
|
170 |
return (generation_a_id, audio_a_path, generation_b_id, audio_b_path)
|
171 |
|
172 |
except Exception as e:
|
173 |
-
if
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
179 |
|
180 |
|
181 |
def parse_hume_tts_generation(generation: Dict[str, Any]) -> Tuple[str, str]:
|
|
|
19 |
"""
|
20 |
|
21 |
# Standard Library Imports
|
|
|
22 |
import logging
|
23 |
+
from dataclasses import dataclass
|
|
|
24 |
from typing import Any, Dict, Literal, Optional, Tuple, Union
|
25 |
|
26 |
# Third-Party Library Imports
|
27 |
import requests
|
28 |
from requests.exceptions import HTTPError
|
29 |
+
from tenacity import after_log, before_log, retry, stop_after_attempt, wait_fixed
|
30 |
|
31 |
# Local Application Imports
|
32 |
+
from src.config import logger, validate_env_var
|
33 |
+
from src.constants import CLIENT_ERROR_CODE, SERVER_ERROR_CODE
|
34 |
+
from src.utils import save_base64_audio_to_file
|
35 |
|
36 |
HumeSupportedFileFormat = Literal["mp3", "pcm", "wav"]
|
37 |
""" Support audio file formats for the Hume TTS API"""
|
|
|
41 |
class HumeConfig:
|
42 |
"""Immutable configuration for interacting with the Hume TTS API."""
|
43 |
|
44 |
+
api_key: Optional[str] = None
|
45 |
url: str = "https://test-api.hume.ai/v0/tts/octave"
|
46 |
headers: dict = None
|
47 |
file_format: HumeSupportedFileFormat = "mp3"
|
|
|
49 |
def __post_init__(self):
|
50 |
# Validate required attributes
|
51 |
if not self.api_key:
|
52 |
+
api_key = validate_env_var("HUME_API_KEY")
|
53 |
+
object.__setattr__(self, "api_key", api_key)
|
54 |
if not self.url:
|
55 |
raise ValueError("Hume TTS endpoint URL is not set.")
|
56 |
if not self.file_format:
|
|
|
117 |
|
118 |
Returns:
|
119 |
Union[Tuple[str, str], Tuple[str, str, str, str]]:
|
120 |
+
- If num_generations == 1: (generation_a_id, audio_a_path).
|
121 |
+
- If num_generations == 2: (generation_a_id, audio_a_path, generation_b_id, audio_b_path).
|
122 |
|
123 |
Raises:
|
124 |
ValueError: If num_generations is not 1 or 2.
|
125 |
HumeError: If there is an error communicating with the Hume TTS API or parsing its response.
|
126 |
UnretryableHumeError: If a client-side HTTP error (status code in the 4xx range) is encountered.
|
127 |
+
Exception: Any other exceptions raised during the request or processing will be wrapped and
|
128 |
+
re-raised as HumeError.
|
129 |
"""
|
130 |
logger.debug(
|
131 |
+
f"Processing TTS with Hume. Prompt length: {len(character_description)} characters. "
|
132 |
+
f"Text length: {len(text)} characters."
|
133 |
)
|
134 |
|
135 |
if num_generations < 1 or num_generations > 2:
|
|
|
171 |
return (generation_a_id, audio_a_path, generation_b_id, audio_b_path)
|
172 |
|
173 |
except Exception as e:
|
174 |
+
if (
|
175 |
+
isinstance(e, HTTPError)
|
176 |
+
and CLIENT_ERROR_CODE <= e.response.status_code < SERVER_ERROR_CODE
|
177 |
+
):
|
178 |
+
raise UnretryableHumeError(
|
179 |
+
message=f"{e.response.text}",
|
180 |
+
original_exception=e,
|
181 |
+
) from e
|
182 |
+
|
183 |
+
raise HumeError(
|
184 |
+
message=f"{e}",
|
185 |
+
original_exception=e,
|
186 |
+
) from e
|
187 |
|
188 |
|
189 |
def parse_hume_tts_generation(generation: Dict[str, Any]) -> Tuple[str, str]:
|
src/utils.py
CHANGED
@@ -8,15 +8,15 @@ These functions provide reusable logic to simplify code in other modules.
|
|
8 |
# Standard Library Imports
|
9 |
import base64
|
10 |
import json
|
11 |
-
import os
|
12 |
import random
|
13 |
import time
|
|
|
14 |
from typing import Tuple
|
15 |
|
16 |
# Local Application Imports
|
17 |
from src import constants
|
18 |
from src.config import AUDIO_DIR, logger
|
19 |
-
from src.
|
20 |
ComparisonType,
|
21 |
Option,
|
22 |
OptionKey,
|
@@ -56,38 +56,6 @@ def truncate_text(text: str, max_length: int = 50) -> str:
|
|
56 |
return text[:max_length] + ("..." if is_truncated else "")
|
57 |
|
58 |
|
59 |
-
def validate_env_var(var_name: str) -> str:
|
60 |
-
"""
|
61 |
-
Validates that an environment variable is set and returns its value.
|
62 |
-
|
63 |
-
Args:
|
64 |
-
var_name (str): The name of the environment variable to validate.
|
65 |
-
|
66 |
-
Returns:
|
67 |
-
str: The value of the environment variable.
|
68 |
-
|
69 |
-
Raises:
|
70 |
-
ValueError: If the environment variable is not set.
|
71 |
-
|
72 |
-
Examples:
|
73 |
-
>>> import os
|
74 |
-
>>> os.environ["EXAMPLE_VAR"] = "example_value"
|
75 |
-
>>> validate_env_var("EXAMPLE_VAR")
|
76 |
-
'example_value'
|
77 |
-
|
78 |
-
>>> validate_env_var("MISSING_VAR")
|
79 |
-
Traceback (most recent call last):
|
80 |
-
...
|
81 |
-
ValueError: MISSING_VAR is not set. Please ensure it is defined in your environment variables.
|
82 |
-
"""
|
83 |
-
value = os.environ.get(var_name, "")
|
84 |
-
if not value:
|
85 |
-
raise ValueError(
|
86 |
-
f"{var_name} is not set. Please ensure it is defined in your environment variables."
|
87 |
-
)
|
88 |
-
return value
|
89 |
-
|
90 |
-
|
91 |
def validate_character_description_length(character_description: str) -> None:
|
92 |
"""
|
93 |
Validates that a voice description is within specified minimum and maximum length limits.
|
@@ -114,16 +82,20 @@ def validate_character_description_length(character_description: str) -> None:
|
|
114 |
|
115 |
if character_description_length < constants.CHARACTER_DESCRIPTION_MIN_LENGTH:
|
116 |
raise ValueError(
|
117 |
-
f"Your character description is too short. Please enter at least
|
|
|
118 |
f"(Current length: {character_description_length})"
|
119 |
)
|
120 |
if character_description_length > constants.CHARACTER_DESCRIPTION_MAX_LENGTH:
|
121 |
raise ValueError(
|
122 |
-
f"Your character description is too long. Please limit it to
|
|
|
123 |
f"(Current length: {character_description_length})"
|
124 |
)
|
|
|
|
|
125 |
logger.debug(
|
126 |
-
f"Character description length validation passed for character_description: {
|
127 |
)
|
128 |
|
129 |
|
@@ -145,28 +117,29 @@ def delete_files_older_than(directory: str, minutes: int = 30) -> None:
|
|
145 |
now = time.time()
|
146 |
# Convert the minutes threshold to seconds.
|
147 |
cutoff = now - (minutes * 60)
|
|
|
148 |
|
149 |
# Iterate over all files in the directory.
|
150 |
-
for
|
151 |
-
file_path
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
|
161 |
|
162 |
def save_base64_audio_to_file(base64_audio: str, filename: str) -> str:
|
163 |
"""
|
164 |
Decode a base64-encoded audio string and write the resulting binary data to a file
|
165 |
within the preconfigured AUDIO_DIR directory. Prior to writing the bytes to an audio
|
166 |
-
file all files within the directory
|
167 |
-
This function verifies the file was created, logs the absolute and relative
|
168 |
-
paths, and returns a path relative to the current working directory
|
169 |
-
Gradio
|
170 |
|
171 |
Args:
|
172 |
base64_audio (str): The base64-encoded string representing the audio data.
|
@@ -178,38 +151,38 @@ def save_base64_audio_to_file(base64_audio: str, filename: str) -> str:
|
|
178 |
str: The relative file path to the saved audio file.
|
179 |
|
180 |
Raises:
|
181 |
-
|
182 |
"""
|
183 |
# Decode the base64-encoded audio into binary data.
|
184 |
audio_bytes = base64.b64decode(base64_audio)
|
185 |
|
186 |
-
# Construct the full absolute file path within the AUDIO_DIR directory.
|
187 |
-
file_path =
|
188 |
|
189 |
-
# Delete all audio files older than 30 minutes before writing new audio file.
|
190 |
num_minutes = 30
|
191 |
delete_files_older_than(AUDIO_DIR, num_minutes)
|
192 |
|
193 |
# Write the binary audio data to the file.
|
194 |
-
with open(
|
195 |
audio_file.write(audio_bytes)
|
196 |
|
197 |
# Verify that the file was created.
|
198 |
-
if not
|
199 |
raise FileNotFoundError(f"Audio file was not created at {file_path}")
|
200 |
|
201 |
-
# Compute a relative path for Gradio to serve (relative to the
|
202 |
-
relative_path =
|
203 |
logger.debug(f"Audio file absolute path: {file_path}")
|
204 |
logger.debug(f"Audio file relative path: {relative_path}")
|
205 |
|
206 |
-
return relative_path
|
207 |
|
208 |
|
209 |
def choose_providers(
|
210 |
text_modified: bool,
|
211 |
character_description: str,
|
212 |
-
) -> Tuple[
|
213 |
"""
|
214 |
Select two TTS providers based on whether the text has been modified.
|
215 |
|
@@ -236,69 +209,48 @@ def choose_providers(
|
|
236 |
else random.choice(constants.TTS_PROVIDERS)
|
237 |
)
|
238 |
|
239 |
-
|
240 |
-
case constants.HUME_AI:
|
241 |
-
comparison_type = constants.HUME_TO_HUME
|
242 |
-
case constants.ELEVENLABS:
|
243 |
-
comparison_type = constants.HUME_TO_ELEVENLABS
|
244 |
-
|
245 |
-
return comparison_type, provider_a, provider_b
|
246 |
|
247 |
|
248 |
-
def create_shuffled_tts_options(
|
249 |
-
provider_a: TTSProviderName,
|
250 |
-
audio_a: str,
|
251 |
-
generation_id_a: str,
|
252 |
-
provider_b: TTSProviderName,
|
253 |
-
audio_b: str,
|
254 |
-
generation_id_b: str,
|
255 |
-
) -> OptionMap:
|
256 |
"""
|
257 |
Create and shuffle TTS generation options.
|
258 |
|
259 |
-
This function
|
260 |
-
|
261 |
-
|
262 |
|
263 |
Args:
|
264 |
-
|
265 |
-
|
266 |
-
generation_id_a (str): The generation ID for the first generation.
|
267 |
-
provider_b (TTSProviderName): The TTS provider for the second generation.
|
268 |
-
audio_b (str): The relative file path to the audio file for the second generation.
|
269 |
-
generation_id_b (str): The generation ID for the second generation.
|
270 |
|
271 |
Returns:
|
272 |
-
|
|
|
273 |
"""
|
274 |
# Create a list of Option instances for the available providers.
|
275 |
-
options = [
|
276 |
-
Option(provider=provider_a, audio=audio_a, generation_id=generation_id_a),
|
277 |
-
Option(provider=provider_b, audio=audio_b, generation_id=generation_id_b),
|
278 |
-
]
|
279 |
|
280 |
# Randomly shuffle the list of options.
|
281 |
random.shuffle(options)
|
282 |
|
283 |
# Unpack the two options.
|
284 |
-
|
285 |
|
286 |
# Build a mapping from option constants to the corresponding providers.
|
287 |
-
|
288 |
"option_a": {
|
289 |
-
"provider":
|
290 |
-
"generation_id":
|
291 |
-
"audio_file_path":
|
292 |
},
|
293 |
"option_b": {
|
294 |
-
"provider":
|
295 |
-
"generation_id":
|
296 |
-
"audio_file_path":
|
297 |
},
|
298 |
}
|
299 |
|
300 |
-
return options_map
|
301 |
-
|
302 |
|
303 |
def determine_selected_option(
|
304 |
selected_option_button: str,
|
@@ -324,11 +276,38 @@ def determine_selected_option(
|
|
324 |
|
325 |
return selected_option, other_option
|
326 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
327 |
|
328 |
def submit_voting_results(
|
329 |
option_map: OptionMap,
|
330 |
selected_option: str,
|
331 |
-
comparison_type: ComparisonType,
|
332 |
text_modified: bool,
|
333 |
character_description: str,
|
334 |
text: str,
|
@@ -347,12 +326,16 @@ def submit_voting_results(
|
|
347 |
Returns:
|
348 |
VotingResults: The constructed voting results dictionary.
|
349 |
"""
|
|
|
|
|
|
|
|
|
350 |
voting_results: VotingResults = {
|
351 |
"comparison_type": comparison_type,
|
352 |
"winning_provider": option_map[selected_option]["provider"],
|
353 |
"winning_option": selected_option,
|
354 |
-
"option_a_provider":
|
355 |
-
"option_b_provider":
|
356 |
"option_a_generation_id": option_map[constants.OPTION_A_KEY]["generation_id"],
|
357 |
"option_b_generation_id": option_map[constants.OPTION_B_KEY]["generation_id"],
|
358 |
"voice_description": character_description,
|
|
|
8 |
# Standard Library Imports
|
9 |
import base64
|
10 |
import json
|
|
|
11 |
import random
|
12 |
import time
|
13 |
+
from pathlib import Path
|
14 |
from typing import Tuple
|
15 |
|
16 |
# Local Application Imports
|
17 |
from src import constants
|
18 |
from src.config import AUDIO_DIR, logger
|
19 |
+
from src.custom_types import (
|
20 |
ComparisonType,
|
21 |
Option,
|
22 |
OptionKey,
|
|
|
56 |
return text[:max_length] + ("..." if is_truncated else "")
|
57 |
|
58 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
def validate_character_description_length(character_description: str) -> None:
|
60 |
"""
|
61 |
Validates that a voice description is within specified minimum and maximum length limits.
|
|
|
82 |
|
83 |
if character_description_length < constants.CHARACTER_DESCRIPTION_MIN_LENGTH:
|
84 |
raise ValueError(
|
85 |
+
f"Your character description is too short. Please enter at least "
|
86 |
+
f"{constants.CHARACTER_DESCRIPTION_MIN_LENGTH} characters. "
|
87 |
f"(Current length: {character_description_length})"
|
88 |
)
|
89 |
if character_description_length > constants.CHARACTER_DESCRIPTION_MAX_LENGTH:
|
90 |
raise ValueError(
|
91 |
+
f"Your character description is too long. Please limit it to "
|
92 |
+
f"{constants.CHARACTER_DESCRIPTION_MAX_LENGTH} characters. "
|
93 |
f"(Current length: {character_description_length})"
|
94 |
)
|
95 |
+
|
96 |
+
truncated_description = truncate_text(stripped_character_description)
|
97 |
logger.debug(
|
98 |
+
f"Character description length validation passed for character_description: {truncated_description}"
|
99 |
)
|
100 |
|
101 |
|
|
|
117 |
now = time.time()
|
118 |
# Convert the minutes threshold to seconds.
|
119 |
cutoff = now - (minutes * 60)
|
120 |
+
dir_path = Path(directory)
|
121 |
|
122 |
# Iterate over all files in the directory.
|
123 |
+
for file_path in dir_path.iterdir():
|
124 |
+
if file_path.is_file():
|
125 |
+
file_mod_time = file_path.stat().st_mtime
|
126 |
+
# If the file's modification time is older than the cutoff, delete it.
|
127 |
+
if file_mod_time < cutoff:
|
128 |
+
try:
|
129 |
+
file_path.unlink()
|
130 |
+
logger.info(f"Deleted: {file_path}")
|
131 |
+
except Exception as e:
|
132 |
+
logger.exception(f"Error deleting {file_path}: {e}")
|
133 |
|
134 |
|
135 |
def save_base64_audio_to_file(base64_audio: str, filename: str) -> str:
|
136 |
"""
|
137 |
Decode a base64-encoded audio string and write the resulting binary data to a file
|
138 |
within the preconfigured AUDIO_DIR directory. Prior to writing the bytes to an audio
|
139 |
+
file, all files within the directory that are more than 30 minutes old are deleted.
|
140 |
+
This function verifies the file was created, logs both the absolute and relative
|
141 |
+
file paths, and returns a path relative to the current working directory
|
142 |
+
(as required by Gradio for serving static files).
|
143 |
|
144 |
Args:
|
145 |
base64_audio (str): The base64-encoded string representing the audio data.
|
|
|
151 |
str: The relative file path to the saved audio file.
|
152 |
|
153 |
Raises:
|
154 |
+
FileNotFoundError: If the audio file was not created.
|
155 |
"""
|
156 |
# Decode the base64-encoded audio into binary data.
|
157 |
audio_bytes = base64.b64decode(base64_audio)
|
158 |
|
159 |
+
# Construct the full absolute file path within the AUDIO_DIR directory using Path.
|
160 |
+
file_path = Path(AUDIO_DIR) / filename
|
161 |
|
162 |
+
# Delete all audio files older than 30 minutes before writing the new audio file.
|
163 |
num_minutes = 30
|
164 |
delete_files_older_than(AUDIO_DIR, num_minutes)
|
165 |
|
166 |
# Write the binary audio data to the file.
|
167 |
+
with file_path.open("wb") as audio_file:
|
168 |
audio_file.write(audio_bytes)
|
169 |
|
170 |
# Verify that the file was created.
|
171 |
+
if not file_path.exists():
|
172 |
raise FileNotFoundError(f"Audio file was not created at {file_path}")
|
173 |
|
174 |
+
# Compute a relative path for Gradio to serve (relative to the current working directory).
|
175 |
+
relative_path = file_path.relative_to(Path.cwd())
|
176 |
logger.debug(f"Audio file absolute path: {file_path}")
|
177 |
logger.debug(f"Audio file relative path: {relative_path}")
|
178 |
|
179 |
+
return str(relative_path)
|
180 |
|
181 |
|
182 |
def choose_providers(
|
183 |
text_modified: bool,
|
184 |
character_description: str,
|
185 |
+
) -> Tuple[TTSProviderName, TTSProviderName]:
|
186 |
"""
|
187 |
Select two TTS providers based on whether the text has been modified.
|
188 |
|
|
|
209 |
else random.choice(constants.TTS_PROVIDERS)
|
210 |
)
|
211 |
|
212 |
+
return provider_a, provider_b
|
|
|
|
|
|
|
|
|
|
|
|
|
213 |
|
214 |
|
215 |
+
def create_shuffled_tts_options(option_a: Option, option_b: Option) -> OptionMap:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
216 |
"""
|
217 |
Create and shuffle TTS generation options.
|
218 |
|
219 |
+
This function accepts two TTS generation options, shuffles them randomly,
|
220 |
+
and returns an OptionMap with keys 'option_a' and 'option_b' corresponding
|
221 |
+
to the shuffled options.
|
222 |
|
223 |
Args:
|
224 |
+
option_a (Option): The first TTS generation option.
|
225 |
+
option_b (Option): The second TTS generation option.
|
|
|
|
|
|
|
|
|
226 |
|
227 |
Returns:
|
228 |
+
OptionMap: A mapping of shuffled TTS options, where each option includes
|
229 |
+
its provider, audio file path, and generation ID.
|
230 |
"""
|
231 |
# Create a list of Option instances for the available providers.
|
232 |
+
options = [option_a, option_b]
|
|
|
|
|
|
|
233 |
|
234 |
# Randomly shuffle the list of options.
|
235 |
random.shuffle(options)
|
236 |
|
237 |
# Unpack the two options.
|
238 |
+
shuffled_option_a, shuffled_option_b = options
|
239 |
|
240 |
# Build a mapping from option constants to the corresponding providers.
|
241 |
+
return {
|
242 |
"option_a": {
|
243 |
+
"provider": shuffled_option_a.provider,
|
244 |
+
"generation_id": shuffled_option_a.generation_id,
|
245 |
+
"audio_file_path": shuffled_option_a.audio,
|
246 |
},
|
247 |
"option_b": {
|
248 |
+
"provider": shuffled_option_b.provider,
|
249 |
+
"generation_id": shuffled_option_b.generation_id,
|
250 |
+
"audio_file_path": shuffled_option_b.audio,
|
251 |
},
|
252 |
}
|
253 |
|
|
|
|
|
254 |
|
255 |
def determine_selected_option(
|
256 |
selected_option_button: str,
|
|
|
276 |
|
277 |
return selected_option, other_option
|
278 |
|
279 |
+
def determine_comparison_type(
|
280 |
+
provider_a: TTSProviderName,
|
281 |
+
provider_b: TTSProviderName
|
282 |
+
) -> ComparisonType:
|
283 |
+
"""
|
284 |
+
Determine the comparison type based on the given TTS provider names.
|
285 |
+
|
286 |
+
If both providers are HUME_AI, the comparison type is HUME_TO_HUME.
|
287 |
+
If either provider is ELEVENLABS, the comparison type is HUME_TO_ELEVENLABS.
|
288 |
+
|
289 |
+
Args:
|
290 |
+
provider_a (TTSProviderName): The first TTS provider.
|
291 |
+
provider_b (TTSProviderName): The second TTS provider.
|
292 |
+
|
293 |
+
Returns:
|
294 |
+
ComparisonType: The determined comparison type.
|
295 |
+
|
296 |
+
Raises:
|
297 |
+
ValueError: If the combination of providers is not recognized.
|
298 |
+
"""
|
299 |
+
if provider_a == constants.HUME_AI and provider_b == constants.HUME_AI:
|
300 |
+
return constants.HUME_TO_HUME
|
301 |
+
|
302 |
+
if constants.ELEVENLABS in (provider_a, provider_b):
|
303 |
+
return constants.HUME_TO_ELEVENLABS
|
304 |
+
|
305 |
+
raise ValueError(f"Invalid provider combination: {provider_a}, {provider_b}")
|
306 |
+
|
307 |
|
308 |
def submit_voting_results(
|
309 |
option_map: OptionMap,
|
310 |
selected_option: str,
|
|
|
311 |
text_modified: bool,
|
312 |
character_description: str,
|
313 |
text: str,
|
|
|
326 |
Returns:
|
327 |
VotingResults: The constructed voting results dictionary.
|
328 |
"""
|
329 |
+
provider_a: TTSProviderName = option_map[constants.OPTION_A_KEY]["provider"]
|
330 |
+
provider_b: TTSProviderName = option_map[constants.OPTION_B_KEY]["provider"]
|
331 |
+
comparison_type: ComparisonType = determine_comparison_type(provider_a, provider_b)
|
332 |
+
|
333 |
voting_results: VotingResults = {
|
334 |
"comparison_type": comparison_type,
|
335 |
"winning_provider": option_map[selected_option]["provider"],
|
336 |
"winning_option": selected_option,
|
337 |
+
"option_a_provider": provider_a,
|
338 |
+
"option_b_provider": provider_b,
|
339 |
"option_a_generation_id": option_map[constants.OPTION_A_KEY]["generation_id"],
|
340 |
"option_b_generation_id": option_map[constants.OPTION_B_KEY]["generation_id"],
|
341 |
"voice_description": character_description,
|