Spaces:
Runtime error
Runtime error
File size: 3,313 Bytes
516059f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 |
from pyharp import ModelCard, build_endpoint, save_and_return_filepath
from audiotools import AudioSignal
from audioldm import build_model, style_transfer
import gradio as gr
import soundfile as sf
audioldm = build_model(model_name="audioldm-m-full")
def save_wave(waveform, savepath, name="outwav"):
if type(name) is not list:
name = [name] * waveform.shape[0]
for i in range(waveform.shape[0]):
path = os.path.join(
savepath,
"%s_%s.wav"
% (
os.path.basename(name[i])
if (not ".wav" in name[i])
else os.path.basename(name[i]).split(".")[0],
i,
),
)
print("Save audio to %s" % path)
sf.write(path, waveform[i, 0], samplerate=16000)
def process_fn(input_audio_path, prompt, seed, guidance_scale, num_inference_steps, num_candidates, audio_length_in_s, transfer_strength):
waveform = style_transfer(
audioldm,
prompt,
audio_file,
transfer_strength,
int(seed),
duration = audio_length_in_s,
guidance_scale = guidance_scale,
ddim_steps = int(num_inference_steps),
batchsize = int(num_candidates),
config=None,
)
waveform = waveform[:,None,:]
sf.write("./output.wav", waveform[0, 0], samplerate=16000)
#save_wave(waveform, "./", name="output.wav") #broken, always appends _0.wav
return "./output.wav"
card = ModelCard(
name='AudioLDM Variations',
description='AudioLDM Variation Generator, operates on region selected in track.',
author='Team Audio',
tags=['AudioLDM', 'Variations', 'audio-to-audio']
)
with gr.Blocks() as webapp:
# Define your Gradio interface
inputs = [
gr.Audio(
label="Audio Input",
type="filepath"
),
gr.Slider(
label="seed",
minimum="0",
maximum="65535",
value="43534",
step="1"
),
gr.Slider(
minimum=0, maximum=10,
step=0.1, value=2.5,
label="Guidance Scale"
),
gr.Slider(
minimum=1, maximum=500,
step=1, value=200,
label="Inference Steps"
),
gr.Slider(
minimum=1, maximum=10,
step=1, value=1,
label="Candidates"
),
gr.Slider(
minimum=2.5, maximum=10.0,
step=2.5, value=5,
label="Duration"
),
]
output = gr.Audio(label="Audio Output", type="filepath")
ctrls_data, ctrls_button, process_button, cancel_button = build_endpoint(inputs, output, process_fn, card)
# queue the webapp: https://www.gradio.app/guides/setting-up-a-demo-for-maximum-performance
#webapp.queue()
webapp.launch(share=True)
for audio_file in input_files:
waveform = style_transfer(
audioldm,
PROMPT,
audio_file,
TRANSFER_STRENGTH,
SEED,
duration = DURATION,
guidance_scale = GUIDANCE_SCALE,
ddim_steps = STEPS,
batchsize = N_VARIATIONS,
config=None,
)
waveform = waveform[:,None,:]
save_wave(waveform, OUTPUT_DIRECTORY, name=f"{os.path.basename(audio_file)}_STYLE_{PROMPT}")
|