jonathanagustin's picture
Upload folder using huggingface_hub
5221577 verified
raw
history blame
4.75 kB
import gradio as gr
import tempfile
import openai
import requests
def tts(input_text: str, model: str, voice: str, api_key: str) -> str:
"""
Convert input text to speech using OpenAI's Text-to-Speech API.
Parameters:
input_text (str): The text to be converted to speech.
model (str): The model to use for synthesis (e.g., 'tts-1', 'tts-1-hd').
voice (str): The voice profile to use (e.g., 'alloy', 'echo', 'fable', etc.).
api_key (str): OpenAI API key.
Returns:
str: File path to the generated audio file.
Raises:
gr.Error: If input parameters are invalid or API call fails.
"""
if not api_key.strip():
raise gr.Error(
"API key is required. Get an API key at: https://platform.openai.com/account/api-keys"
)
if not input_text.strip():
raise gr.Error("Input text cannot be empty.")
openai.api_key = api_key
try:
response = openai.Audio.create(text=input_text, voice=voice, model=model)
except openai.OpenAIError as e:
# Catch-all for OpenAI exceptions
raise gr.Error(f"An OpenAI error occurred: {e}")
except Exception as e:
# Catch any other exceptions
raise gr.Error(f"An unexpected error occurred: {e}")
if not hasattr(response, "audio"):
raise gr.Error(
"Invalid response from OpenAI API. The response does not contain audio content."
)
with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as temp_file:
temp_file.write(response.audio)
temp_file_path = temp_file.name
return temp_file_path
def main():
"""
Main function to create and launch the Gradio interface.
"""
MODEL_OPTIONS = ["tts-1", "tts-1-hd"]
VOICE_OPTIONS = ["alloy", "echo", "fable", "onyx", "nova", "shimmer"]
# Predefine voice previews URLs
VOICE_PREVIEWS = {
voice: f"https://cdn.openai.com/API/docs/audio/{voice}.wav"
for voice in VOICE_OPTIONS
}
with gr.Blocks() as demo:
with gr.Row():
with gr.Column(scale=1):
api_key_input = gr.Textbox(
label="OpenAI API Key",
info="https://platform.openai.com/account/api-keys",
type="password",
placeholder="Enter your OpenAI API Key",
)
model_dropdown = gr.Dropdown(
choices=MODEL_OPTIONS, label="Model", value="tts-1"
)
voice_dropdown = gr.Dropdown(
choices=VOICE_OPTIONS, label="Voice Options", value="echo"
)
# Add voice previews
gr.Markdown("### Voice Previews")
for voice in VOICE_OPTIONS:
audio_url = VOICE_PREVIEWS[voice]
# Fetch the audio data
try:
response = requests.get(audio_url)
response.raise_for_status()
audio_data = response.content
gr.Audio(
value=audio_data,
waveform_options=gr.WaveformOptions(
show_download_button=False,
),
show_share_button=False,
show_controls=False,
label=f"{voice.capitalize()}",
autoplay=False,
)
except requests.exceptions.RequestException as e:
gr.Markdown(
f"Could not load preview for {voice.capitalize()}: {e}"
)
with gr.Column(scale=2):
input_textbox = gr.Textbox(
label="Input Text", lines=10, placeholder="Type your text here..."
)
submit_button = gr.Button("Convert Text to Speech", variant="primary")
with gr.Column(scale=1):
output_audio = gr.Audio(label="Output Audio")
# Define the event handler for the submit button with error handling
def on_submit(input_text, model, voice, api_key):
audio_file = tts(input_text, model, voice, api_key)
return audio_file
# Trigger the conversion when the submit button is clicked
submit_button.click(
fn=on_submit,
inputs=[input_textbox, model_dropdown, voice_dropdown, api_key_input],
outputs=output_audio,
)
# Launch the Gradio app with error display enabled
demo.launch(show_error=True)
if __name__ == "__main__":
main()