Format choice

app.py

@@ -7,6 +7,7 @@ from huggingface_hub import snapshot_download
 from models import AudioDiffusion, DDPMScheduler
 from audioldm.audio.stft import TacotronSTFT
 from audioldm.variational_autoencoder import AutoencoderKL
+from pydub import AudioSegment
 from gradio import Markdown
 import spaces
 
@@ -83,18 +84,24 @@ tango.stft.to(device_type)
 tango.model.to(device_type)
     
 @spaces.GPU(duration=120)
-def gradio_generate(prompt, steps, guidance):
+def gradio_generate(prompt, output_format, steps, guidance):
     output_wave = tango.generate(prompt, steps, guidance)
     # output_filename = f"{prompt.replace(' ', '_')}_{steps}_{guidance}"[:250] + ".wav"
 
-    output_filename_1 = "tmp1_.wav"
-    wavio.write(output_filename_1, output_wave[0], rate=16000, sampwidth=2)
+    output_filename_1 = "tmp1.wav"
+    wavio.write(output_filename_1, output_wave, rate=16000, sampwidth=2)
+    output_filename_2 = "tmp2.wav"
+    wavio.write(output_filename_2, output_wave, rate=16000, sampwidth=2)
+    output_filename_3 = "tmp3.wav"
+    wavio.write(output_filename_3, output_wave, rate=16000, sampwidth=2)
 
-    output_filename_2 = "tmp2_.wav"
-    wavio.write(output_filename_2, output_wave[1], rate=16000, sampwidth=2)
-
-    output_filename_3 = "tmp3_.wav"
-    wavio.write(output_filename_3, output_wave[2], rate=16000, sampwidth=2)
+    if (output_format == "mp3"):
+        AudioSegment.from_wav("tmp1.wav").export("tmp1.mp3", format = "mp3")
+        output_filename_1 = "tmp1.mp3"
+        AudioSegment.from_wav("tmp2.wav").export("tmp2.mp3", format = "mp3")
+        output_filename_2 = "tmp2.mp3"
+        AudioSegment.from_wav("tmp3.wav").export("tmp3.mp3", format = "mp3")
+        output_filename_3 = "tmp3.mp3"
 
     return [output_filename_1, output_filename_2, output_filename_3]
 
@@ -118,13 +125,14 @@ def gradio_generate(prompt, steps, guidance):
 # <p/>
 # """
 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/>
+<p><a href="https://huggingface.co/spaces/declare-lab/tango2-full/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 download", choices = ["mp3", "wav"], value = "wav")
 output_audio_1 = gr.Audio(label="Generated Audio #1/3", type="filepath")
 output_audio_2 = gr.Audio(label="Generated Audio #2/3", type="filepath")
 output_audio_3 = gr.Audio(label="Generated Audio #3/3", type="filepath")
@@ -134,7 +142,7 @@ guidance_scale = gr.Slider(minimum=1, maximum=10, value=3, step=0.1, label="Guid
 # Gradio interface
 gr_interface = gr.Interface(
     fn=gradio_generate,
-    inputs=[input_text, denoising_steps, guidance_scale],
+    inputs=[input_text, output_format, denoising_steps, guidance_scale],
     outputs=[output_audio_1, output_audio_2, output_audio_3],
     title="Tango 2: Aligning Diffusion-based Text-to-Audio Generations through Direct Preference Optimization",
     description=description_text,