mp3 or wav

app.py

@@ -2,12 +2,14 @@ import gradio as gr
 import json
 import torch
 import time
+import wavio
 
 from tqdm import tqdm
 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
 
 # Automatic device detection
 if torch.cuda.is_available():
@@ -82,29 +84,61 @@ tango.model.to(device_type)
 
 def check(
     prompt,
+    output_format,
     output_number,
     steps,
     guidance
 ):
     if prompt is None or prompt == "":
         raise gr.Error("Please provide a prompt input.")
+    if not output_format in ["wav", "mp3"]:
+        raise gr.Error("Please choose an allowed output format (.wav or .mp3).")
     if not output_number in [1, 2, 3]:
         raise gr.Error("Please ask for 1, 2 or 3 output files.")
 
-def update_display(output_number):
-    return [gr.update(visible = (2 <= output_number)), gr.update(visible = (output_number == 3))]
+def update_output(output_format, output_number):
+    return [
+        gr.update(format = output_format),
+        gr.update(format = output_format, visible = (2 <= output_number)),
+        gr.update(format = output_format, visible = (output_number == 3))
+    ]
 
 def text2audio(
     prompt,
+    output_format,
     output_number,
     steps,
     guidance
 ):
     start = time.time()
     output_wave = tango.generate(prompt, steps, guidance, output_number)
-    output_wave_1 = gr.make_waveform((16000, output_wave[0]))
-    output_wave_2 = gr.make_waveform((16000, output_wave[1])) if (2 <= output_number) else None
-    output_wave_3 = gr.make_waveform((16000, output_wave[2])) if (output_number == 3) else None
+
+    output_filename_1 = "tmp1.wav"
+    wavio.write(output_filename_1, output_wave[0], rate = 16000, sampwidth = 2)
+
+    if (output_format == "mp3"):
+        AudioSegment.from_wav("tmp1.wav").export("tmp1.mp3", format = "mp3")
+        output_filename_1 = "tmp1.mp3"
+
+    if (2 <= output_number):
+        output_filename_2 = "tmp2.wav"
+        wavio.write(output_filename_2, output_wave[1], rate = 16000, sampwidth = 2)
+
+        if (output_format == "mp3"):
+            AudioSegment.from_wav("tmp2.wav").export("tmp2.mp3", format = "mp3")
+            output_filename_2 = "tmp2.mp3"
+    else:
+        output_filename_2 = None
+
+    if (output_number == 3):
+        output_filename_3 = "tmp3.wav"
+        wavio.write(output_filename_3, output_wave[2], rate = 16000, sampwidth = 2)
+
+        if (output_format == "mp3"):
+            AudioSegment.from_wav("tmp3.wav").export("tmp3.mp3", format = "mp3")
+            output_filename_3 = "tmp3.mp3"
+    else:
+        output_filename_3 = None
 
     end = time.time()
     secondes = int(end - start)
@@ -113,10 +147,10 @@ def text2audio(
     hours = minutes // 60
     minutes = minutes - (hours * 60)
     return [
-        output_wave_1,
-        output_wave_2,
-        output_wave_3,
-        "Start again to get a different result. The output have been generated in " + str(hours) + " h, " + str(minutes) + " min, " + str(secondes) + " sec."
+        output_filename_1,
+        output_filename_2,
+        output_filename_3,
+        "Start again to get a different result. The output have been generated in " + ((str(hours) + " h, ") if hours != 0 else "") + ((str(minutes) + " min, ") if hours != 0 or minutes != 0 else "") + str(secondes) + " sec."
     ]
 
 # Gradio interface
@@ -141,30 +175,35 @@ with gr.Blocks() as interface:
         """
     )
     input_text = gr.Textbox(label = "Prompt", value = "Snort of a horse", lines = 2, autofocus = True)
+    output_format = gr.Radio(label = "Output format", info = "The file you can dowload", choices = ["mp3", "wav"], value = "wav")
     with gr.Accordion("Advanced options", open = False):
         output_number = gr.Slider(label = "Number of generations", info = "1, 2 or 3 output files", minimum = 1, maximum = 3, value = 3, step = 1, interactive = True)
         denoising_steps = gr.Slider(label = "Steps", info = "lower=faster & variant, higher=audio quality & similar", minimum = 100, maximum = 200, value = 100, step = 1, interactive = True)
         guidance_scale = gr.Slider(label = "Guidance Scale", info = "lower=audio quality, higher=follow the prompt", minimum = 1, maximum = 10, value = 3, step = 0.1, interactive = True)
 
-    submit = gr.Button("Generate 🚀", variant = "primary")
+    submit = gr.Button("🚀 Generate", variant = "primary")
 
-    output_audio_1 = gr.Audio(label = "Generated Audio #1/3")
-    output_audio_2 = gr.Audio(label = "Generated Audio #2/3")
-    output_audio_3 = gr.Audio(label = "Generated Audio #3/3")
+    output_audio_1 = gr.Audio(label = "Generated Audio #1/3", format = "wav", type="filepath", autoplay = True)
+    output_audio_2 = gr.Audio(label = "Generated Audio #2/3", format = "wav", type="filepath")
+    output_audio_3 = gr.Audio(label = "Generated Audio #3/3", format = "wav", type="filepath")
     information = gr.Label(label = "Information")
 
     submit.click(fn = check, inputs = [
         input_text,
+        output_format,
         output_number,
         denoising_steps,
         guidance_scale
-    ], outputs = [], queue = False, show_progress = False).success(fn = update_display, inputs = [
+    ], outputs = [], queue = False, show_progress = False).success(fn = update_output, inputs = [
+        output_format,
         output_number
     ], outputs = [
+        output_audio_1,
         output_audio_2,
         output_audio_3
     ], queue = False, show_progress = False).success(fn = text2audio, inputs = [
         input_text,
+        output_format,
         output_number,
         denoising_steps,
         guidance_scale
@@ -179,6 +218,7 @@ with gr.Blocks() as interface:
         fn = text2audio,
 	    inputs = [
             input_text,
+            output_format,
             output_number,
             denoising_steps,
             guidance_scale
@@ -190,11 +230,11 @@ with gr.Blocks() as interface:
             information
         ],
         examples = [
-                ["A hammer is hitting a wooden surface", 3, 100, 3],
-                ["Peaceful and calming ambient music with singing bowl and other instruments.", 3, 100, 3],
-                ["A man is speaking in a small room.", 3, 100, 3],
-                ["A female is speaking followed by footstep sound", 3, 100, 3],
-                ["Wooden table tapping sound followed by water pouring sound.", 3, 100, 3],
+                ["A hammer is hitting a wooden surface", "mp3", 3, 100, 3],
+                ["Peaceful and calming ambient music with singing bowl and other instruments.", "wav", 3, 100, 3],
+                ["A man is speaking in a small room.", "mp3", 2, 100, 3],
+                ["A female is speaking followed by footstep sound", "mp3", 1, 100, 3],
+                ["Wooden table tapping sound followed by water pouring sound.", "mp3", 3, 200, 3],
             ],
         cache_examples = "lazy",
     )