Spaces:
Paused
Paused
| import os | |
| import gradio as gr | |
| from langchain import PromptTemplate | |
| from langchain.chains import LLMChain | |
| from langchain.llms import OpenAI | |
| from gradio_client import Client | |
| eleven = Client("https://elevenlabs-tts.hf.space/") | |
| import wave | |
| openai_api_key = os.environ.get("OPENAI_API_KEY") | |
| llm = OpenAI(temperature=0.9) | |
| def split_text_into_sentences(text): | |
| sentences = [] | |
| current_sentence = '' | |
| words = text.split() | |
| for word in words: | |
| current_sentence += ' ' + word | |
| if word.endswith('.'): | |
| sentences.append(current_sentence.strip()) | |
| current_sentence = '' | |
| if current_sentence: | |
| sentences.append(current_sentence.strip()) | |
| return sentences | |
| def join_wav_files(input_files, output_file): | |
| # Open the first input file to get its parameters | |
| with wave.open(input_files[0], 'rb') as first_file: | |
| # Get the audio parameters from the first file | |
| params = first_file.getparams() | |
| # Create a new wave file for writing the joined audio | |
| with wave.open(output_file, 'wb') as output: | |
| output.setparams(params) | |
| # Iterate over the input files and write their audio data to the output file | |
| for input_file in input_files: | |
| with wave.open(input_file, 'rb') as input: | |
| output.writeframes(input.readframes(input.getnframes())) | |
| def generate_story(text): | |
| """Generate a story using the langchain library and OpenAI's GPT-3 model.""" | |
| prompt = PromptTemplate( | |
| input_variables=["text"], | |
| template=""" | |
| You are a fun and seasoned storyteller. | |
| Generate a short bedtime story for a 5 years old audience about {text}. | |
| Use short sentences. The story is not too long, but not too short either. | |
| Always finish your story with "The End". | |
| """ | |
| ) | |
| story = LLMChain(llm=llm, prompt=prompt) | |
| story_result = story.run(text=text) | |
| print(story_result) | |
| print(""" | |
| — | |
| Cutting text in chunks | |
| — | |
| """) | |
| input_waves = [] | |
| text_chunks = split_text_into_sentences(story_result) | |
| for chunk in text_chunks: | |
| print(chunk) | |
| result = eleven.predict( | |
| chunk, # str representing input in 'Input Text (250 characters max)' Textbox component | |
| "Bella", # str representing input in 'Voice' Dropdown component | |
| "eleven_monolingual_v1", # str representing input in 'Model' Radio component | |
| fn_index=0 | |
| ) | |
| print(result) | |
| input_waves.append(result) | |
| output_wav = 'output.wav' | |
| join_wav_files(input_waves, output_wav) | |
| return story_result, 'output.wav' | |
| def app(text): | |
| story = generate_story(text) | |
| return story | |
| css = """ | |
| #col-container {max-width: 510px; margin-left: auto; margin-right: auto;} | |
| a {text-decoration-line: underline; font-weight: 600;} | |
| """ | |
| with gr.Blocks(css=css) as demo: | |
| with gr.Column(elem_id="col-container"): | |
| text = gr.Textbox(label="Subject", info="Will generate and tell a story about {your text input}") | |
| submit_btn = gr.Button('Submit') | |
| audio = gr.Audio("The story audio told") | |
| story = gr.Textbox(label="The story text") | |
| submit_btn.click(fn=app, inputs=[text], outputs=[story, audio]) | |
| demo.launch() |