Added n_candidates

app.py

@@ -19,7 +19,7 @@ processor = AutoProcessor.from_pretrained("sanchit-gandhi/clap-htsat-unfused-m-f
 
 generator = torch.Generator(device)
 
-def texttoaudio(prompt, neg_prompt, seed, inf_steps, guidance_scale):
+def texttoaudio(prompt, neg_prompt, seed, inf_steps, guidance_scale, n_candidates):
     if prompt is None:
         raise gr.Error("Please provide a text input.")
         
@@ -30,6 +30,7 @@ def texttoaudio(prompt, neg_prompt, seed, inf_steps, guidance_scale):
                  guidance_scale=guidance_scale,
                  audio_length_in_s=5.0,
                  generator=generator.manual_seed(int(seed)),
+                 num_waveforms_per_prompt=n_candidates if n_candidates else 1,
                 )["audios"]
 
     # save the audio sample as a .wav file
@@ -51,6 +52,6 @@ def score_waveforms(text, waveforms):
     waveform = waveforms[most_probable]
     return waveform
 
-iface = gr.Interface(fn=texttoaudio, title="Prompt, Neg Prompt, Seed, Inf Steps, Guidance Scale", inputs=["text", "text", "number", "number", "number"], outputs="audio")
+iface = gr.Interface(fn=texttoaudio, title="AudioLDM Testing Playground", inputs=["text", "text", "number", "number", "number", "number"], outputs="audio")
 
 iface.launch()