app.py

import spaces
import gradio as gr
import json
import torch
import wavio
from tqdm import tqdm
from huggingface_hub import snapshot_download
from pydub import AudioSegment
from gradio import Markdown
import uuid
import torch
from diffusers import DiffusionPipeline,AudioPipelineOutput
from transformers import CLIPTextModel, T5EncoderModel, AutoModel, T5Tokenizer, T5TokenizerFast
from typing import Union
from diffusers.utils.torch_utils import randn_tensor
from tqdm import tqdm
from TangoFlux import TangoFluxInference
import torchaudio



tangoflux = TangoFluxInference(name="declare-lab/TangoFlux")



@spaces.GPU(duration=15)
def gradio_generate(prompt, steps, guidance,duration=10):

    output = tangoflux.generate(prompt,steps=steps,guidance_scale=guidance,duration=duration)
    #output = output.to(torch.float32).div(torch.max(torch.abs(output))).clamp(-1, 1).mul(32767).to(torch.int16).cpu()
    
    
    #wavio.write(output_filename, output_wave, rate=44100, sampwidth=2)
    filename = 'temp.wav'
    #print(f"Saving audio to file: {unique_filename}")

    # Save to file
    torchaudio.save(filename, output, 44100)
   

    # Return the path to the generated audio file
    return filename

    #if (output_format == "mp3"):
     #   AudioSegment.from_wav("temp.wav").export("temp.mp3", format = "mp3")
      #  output_filename = "temp.mp3"

    #return output_filename

description_text = """
<p><a href="https://huggingface.co/spaces/declare-lab/tango2/blob/main/app.py?duplicate=true"> <img style="margin-top: 0em; margin-bottom: 0em" src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a> For faster inference without waiting in queue, you may duplicate the space and upgrade to a GPU in the settings. <br/><br/>
Generate audio using Tango2 by providing a text prompt. Tango2 was built from Tango and was trained on <a href="https://huggingface.co/datasets/declare-lab/audio-alpaca">Audio-alpaca</a>
<br/><br/> This is the demo for Tango2 for text to audio generation: <a href="https://arxiv.org/abs/2404.09956">Read our paper.</a>
<p/>
"""
# Gradio input and output components
input_text = gr.Textbox(lines=2, label="Prompt")
#output_format = gr.Radio(label = "Output format", info = "The file you can dowload", choices =  "wav"], value = "wav")
output_audio = gr.Audio(label="Generated Audio", type="filepath")
denoising_steps = gr.Slider(minimum=10, maximum=100, value=25, step=5, label="Steps", interactive=True)
guidance_scale = gr.Slider(minimum=1, maximum=10, value=4.5, step=0.5, label="Guidance Scale", interactive=True)
duration_scale = gr.Slider(minimum=1, maximum=30, value=10, step=1, label="Duration", interactive=True)


# Gradio interface
gr_interface = gr.Interface(
    fn=gradio_generate,
    inputs=[input_text, denoising_steps, guidance_scale,duration_scale],
    outputs=output_audio,
    title="TangoFlux: ",
    description=description_text,
    allow_flagging=False,




    examples=[
["Quiet whispered conversation gradually fading into distant jet engine roar diminishing into silence"],
["Clear sound of bicycle tires crunching on loose gravel and dirt, followed by deep male laughter echoing"],
["Multiple ducks quacking loudly with splashing water and piercing wild animal shriek in background"],
["Powerful ocean waves crashing and receding on sandy beach with distant seagulls"],
["Gentle female voice cooing and baby responding with happy gurgles and giggles"],
["Clear male voice speaking, sharp popping sound, followed by genuine group laughter"],
["Stream of water hitting empty ceramic cup, pitch rising as cup fills up"],
["Massive crowd erupting in thunderous applause and excited cheering"],
["Deep rolling thunder with bright lightning strikes crackling through sky"],
["Aggressive dog barking and distressed cat meowing as racing car roars past at high speed"],
["Peaceful stream bubbling and birds singing, interrupted by sudden explosive gunshot"],
["Man speaking outdoors, goat bleating loudly, metal gate scraping closed, ducks quacking frantically, wind howling into microphone"],
["Series of loud aggressive dog barks echoing"],
["Multiple distinct cat meows at different pitches"],
["Rhythmic wooden table tapping overlaid with steady water pouring sound"],
["Sustained crowd applause with camera clicks and amplified male announcer voice"],
["Two sharp gunshots followed by panicked birds taking flight with rapid wing flaps"],
["Melodic human whistling harmonizing with natural birdsong"],
["Deep rhythmic snoring with clear breathing patterns"],
["Multiple racing engines revving and accelerating with sharp whistle piercing through"],
["Massive stadium crowd cheering as thunder crashes and lightning strikes"],
["Heavy helicopter blades chopping through air with engine and wind noise"],
["Dog barking excitedly and man shouting as race car engine roars past"]
    ],
    cache_examples="lazy", # Turn on to cache.
)



gr_interface.queue(15).launch()