Bugfix
Browse filesCasted inf_steps to an integer, and changed the audio variable resulting from the pipe call into what AudioLDM writes in their code
app.py
CHANGED
@@ -18,15 +18,15 @@ def texttoaudio(prompt, neg_prompt, seed, inf_steps, guidance_scale):
|
|
18 |
audio = pipe(
|
19 |
prompt,
|
20 |
negative_prompt=neg_prompt,
|
21 |
-
num_inference_steps=inf_steps,
|
22 |
guidance_scale=guidance_scale,
|
23 |
audio_length_in_s=5.0,
|
24 |
generator=generator.manual_seed(int(seed)),
|
25 |
-
)
|
26 |
|
27 |
# save the audio sample as a .wav file
|
28 |
# scipy.io.wavfile.write("output.wav", rate=16000, data=audio)
|
29 |
-
return (16000, audio)
|
30 |
|
31 |
iface = gr.Interface(fn=texttoaudio, title="Prompt, Neg Prompt, Seed, Inf Steps, Guidance Scale", inputs=["text", "text", "number", "number", "number"], outputs="audio")
|
32 |
|
|
|
18 |
audio = pipe(
|
19 |
prompt,
|
20 |
negative_prompt=neg_prompt,
|
21 |
+
num_inference_steps=int(inf_steps),
|
22 |
guidance_scale=guidance_scale,
|
23 |
audio_length_in_s=5.0,
|
24 |
generator=generator.manual_seed(int(seed)),
|
25 |
+
)["audios"]
|
26 |
|
27 |
# save the audio sample as a .wav file
|
28 |
# scipy.io.wavfile.write("output.wav", rate=16000, data=audio)
|
29 |
+
return (16000, audio[0])
|
30 |
|
31 |
iface = gr.Interface(fn=texttoaudio, title="Prompt, Neg Prompt, Seed, Inf Steps, Guidance Scale", inputs=["text", "text", "number", "number", "number"], outputs="audio")
|
32 |
|