app.py

import os
import gradio as gr
from inspiremusic.cli.inference import InspireMusicUnified, set_env_variables

def get_args(
		task, text="", audio=None, model_name="InspireMusic-Base",
		chorus="intro", fast=False, fade_out=True, trim=False,
		output_sample_rate=48000, max_generate_audio_seconds=30.0):
	# This function constructs the arguments required for InspireMusic
	args = {
		"task"                      : task,
		"text"                      : text,
		"audio_prompt"              : audio,
		"model_name"                : model_name,
		"chorus"                    : chorus,
		"fast"                      : fast,
		"fade_out"                  : fade_out,
		"trim"                      : trim,
		"output_sample_rate"        : output_sample_rate,
		"max_generate_audio_seconds": max_generate_audio_seconds,
		"model_dir"                 : os.path.join("pretrained_models",
												   model_name),
		"result_dir"                : "exp/inspiremusic",
		"output_fn"                 : "output_audio",
		"format"                    : "wav",
	}
	return args


def music_generation(args):
	set_env_variables()
	model = InspireMusicUnified(
			model_name=args["model_name"],
			model_dir=args["model_dir"],
			fast=args["fast"],
			fade_out_mode=args["fade_out"],
			trim=args["trim"],
			output_sample_rate=args["output_sample_rate"],
			max_generate_audio_seconds=args["max_generate_audio_seconds"]
	)
	output_path = model.inference(
			task=args["task"],
			text=args["text"],
			audio_prompt=args["audio_prompt"],
			chorus=args["chorus"],
			output_fn=args["output_fn"],
			output_format=args["format"],
			time_start=0.0, time_end=30.0
	)
	return output_path


def run_inspiremusic(task, text, audio, model_name, chorus, fast, fade_out,
					 trim, output_sample_rate, max_generate_audio_seconds):
	args = get_args(
			task=task, text=text, audio=audio,
			model_name=model_name, chorus=chorus, fast=fast,
			fade_out=fade_out, trim=trim, output_sample_rate=output_sample_rate,
			max_generate_audio_seconds=max_generate_audio_seconds)
	return music_generation(args)


with gr.Blocks() as demo:
	gr.Markdown("""
    # InspireMusic:
    Generate music using InspireMusic with various tasks such as "Text-to-Music" or "Music Continuation".
    """)

	with gr.Row():
		task = gr.Radio(["text-to-music", "continuation"], label="Select Task",
						value="text-to-music")
		model_name = gr.Dropdown(["InspireMusic-1.5B-Long", "InspireMusic-1.5B", "InspireMusic-1.5B-24kHz", "InspireMusic-Base", "InspireMusic-Base-24kHz"], label="Model Name", value="InspireMusic-Base")

	text_input = gr.Textbox(label="Input Text (For Text-to-Music Task)")
	audio_input = gr.Audio(label="Input Audio (For Music Continuation Task)",
						   source="upload", type="filepath")

	with gr.Column():
		chorus = gr.Dropdown(["verse", "chorus", "intro", "outro"],
							 label="Chorus Mode", value="intro")
		fast = gr.Checkbox(label="Fast Inference", value=False)
		fade_out = gr.Checkbox(label="Apply Fade Out", value=True)
		trim = gr.Checkbox(label="Trim Silence", value=False)
		output_sample_rate = gr.Dropdown([24000, 48000],
										 label="Output Sample Rate",
										 value=48000)
		max_generate_audio_seconds = gr.Slider(10, 300,
											   label="Max Generated Audio Length (Seconds)",
											   value=30)

	music_output = gr.Audio(label="Generated Music Result", type="filepath")
	generate_button = gr.Button("Generate Music")

	generate_button.click(run_inspiremusic,
						  inputs=[task, text_input, audio_input, model_name,
								  chorus, fast, fade_out, trim,
								  output_sample_rate,
								  max_generate_audio_seconds],
						  outputs=music_output)

demo.launch()