Write outputs

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import json
 import torch
 import time
+import wavio
 
 from tqdm import tqdm
 from huggingface_hub import snapshot_download
@@ -65,7 +66,7 @@ class Tango:
         for k in tqdm(range(0, len(prompts), batch_size)):
             batch = prompts[k: k + batch_size]
             with torch.no_grad():
-                latents = self.model.inference(batch, self.scheduler, steps, guidance, samples, disable_progress = disable_progress)
+                latents = self.model.inference(batch, self.scheduler, steps, guidance, samples, disable_progress = disable_progress, length = 20)
                 mel = self.vae.decode_first_stage(latents)
                 wave = self.vae.decode_to_waveform(mel)
                 outputs += [item for item in wave]
@@ -110,9 +111,21 @@ def text2audio(
 ):
     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, output_wave[0], rate = 16000, sampwidth = 2)
+
+    if (2 <= output_number):
+        output_filename_2 = "tmp2_.wav"
+        wavio.write(output_filename, output_wave[1], rate = 16000, sampwidth = 2)
+    else:
+        output_filename_2 = None
+
+    if (output_number == 3):
+        output_filename_3 = "tmp3_.wav"
+        wavio.write(output_filename, output_wave[2], rate = 16000, sampwidth = 2)
+    else:
+        output_filename_3 = None
 
     end = time.time()
     secondes = int(end - start)
@@ -121,10 +134,10 @@ 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
@@ -149,7 +162,7 @@ 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 = "mp3")
+    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)
@@ -157,9 +170,9 @@ with gr.Blocks() as interface:
 
     submit = gr.Button("🚀 Generate", variant = "primary")
 
-    output_audio_1 = gr.Audio(label = "Generated Audio #1/3", format = "mp3", autoplay = True)
-    output_audio_2 = gr.Audio(label = "Generated Audio #2/3", format = "mp3")
-    output_audio_3 = gr.Audio(label = "Generated Audio #3/3", format = "mp3")
+    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 = [