3 output files

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 import json
 import torch
+import time
 
 from tqdm import tqdm
 from huggingface_hub import snapshot_download
@@ -56,7 +57,7 @@ class Tango:
             latents = self.model.inference([prompt], self.scheduler, steps, guidance, samples, disable_progress = disable_progress)
             mel = self.vae.decode_first_stage(latents)
             wave = self.vae.decode_to_waveform(mel)
-        return wave[0]
+        return wave
     
     def generate_for_batch(self, prompts, steps = 200, guidance = 3, samples = 1, batch_size = 8, disable_progress = True):
         # Generate audio for a list of prompt strings
@@ -81,26 +82,49 @@ tango.model.to(device_type)
 
 def check(
     prompt,
+    output_number,
     steps,
     guidance
 ):
     if prompt is None or prompt == "":
         raise gr.Error("Please provide a prompt input.")
+    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 text2audio(
     prompt,
+    output_number,
     steps,
     guidance
 ):
-    output_wave = tango.generate(prompt, steps, guidance)
-    return gr.make_waveform((16000, output_wave))
+    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
+
+    end = time.time()
+    secondes = int(end - start)
+    minutes = secondes // 60
+    secondes = secondes - (minutes * 60)
+    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."
+    ]
 
 # Gradio interface
 with gr.Blocks() as interface:
     gr.Markdown("""
         <p style="text-align: center;">
         <b><big><big><big>Text-to-Audio</big></big></big></b>
-        <br/>Generates an audio file, freely, without account, without watermark, that you can download.
+        <br/>Generates 10 seconds of sound effects from description, freely, without account, without watermark, that you can download.
         </p>
         <br/>
         <br/>
@@ -110,7 +134,7 @@ with gr.Blocks() as interface:
         <li>If you need to generate <b>music</b>, I recommend to use <i>MusicGen</i>,</li>
         </ul>
         <br/>
-        🐌 Slow process... Your computer must <b><u>not</u></b> enter into standby mode.<br/>You can duplicate this space on a free account, it works on CPU.<br/>
+        🐌 Slow process... ~2 hours. Your computer must <b><u>not</u></b> enter into standby mode.<br/>You can duplicate this space on a free account, it works on CPU.<br/>
         <a href='https://huggingface.co/spaces/Fabrice-TIERCELIN/Text-to-Audio?duplicate=true'><img src='https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14'></a>
         <br/>
         ⚖️ You can use, modify and share the generated sounds but not for commercial uses.
@@ -118,43 +142,82 @@ with gr.Blocks() as interface:
     )
     input_text = gr.Textbox(label = "Prompt", value = "Snort of a horse", lines = 2, autofocus = True)
     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")
 
-    output_audio = gr.Audio(label = "Generated Audio")
+    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")
+    information = gr.Label(label = "Information")
 
     submit.click(fn = check, inputs = [
         input_text,
+        output_number,
         denoising_steps,
         guidance_scale
-    ], outputs = [], queue = False, show_progress = False).success(fn = text2audio, inputs = [
+    ], outputs = [], queue = False, show_progress = False).success(fn = update_display, inputs = [
+        output_number
+    ], outputs = [
+        output_audio_2,
+        output_audio_3
+    ], queue = False, show_progress = False).success(fn = text2audio, inputs = [
         input_text,
+        output_number,
         denoising_steps,
         guidance_scale
     ], outputs = [
-        output_audio
+        output_audio_1,
+        output_audio_2,
+        output_audio_3,
+        information
     ], scroll_to_output = True)
 
     gr.Examples(
         fn = text2audio,
 	    inputs = [
             input_text,
+            output_number,
             denoising_steps,
             guidance_scale
         ],
 	    outputs = [
-            output_audio
+            output_audio_1,
+            output_audio_2,
+            output_audio_3,
+            information
         ],
         examples = [
-                ["A hammer is hitting a wooden surface", 100, 3],
-                ["Peaceful and calming ambient music with singing bowl and other instruments.", 100, 3],
-                ["A man is speaking in a small room.", 100, 3],
-                ["A female is speaking followed by footstep sound", 100, 3],
-                ["Wooden table tapping sound followed by water pouring sound.", 100, 3],
+                ["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],
             ],
         cache_examples = "lazy",
     )
+    
+    gr.Markdown(
+        """
+        ## How to prompt your sound
+        You can use round brackets to increase the importance of a part:
+        ```
+        Peaceful and (calming) ambient music with singing bowl and other instruments
+        ```
+        You can use several levels of round brackets to even more increase the importance of a part:
+        ```
+        (Peaceful) and ((calming)) ambient music with singing bowl and other instruments
+        ```
+        You can use number instead of several round brackets:
+        ```
+        (Peaceful:1.5) and ((calming)) ambient music with singing bowl and other instruments
+        ```
+        You can do the same thing with square brackets to decrease the importance of a part:
+        ```
+        (Peaceful:1.5) and ((calming)) ambient music with [singing:2] bowl and other instruments
+        """
+    )
         
     interface.queue(10).launch()