Spaces:
Running
on
Zero
Running
on
Zero
changed back to dropdown
Browse files
app.py
CHANGED
@@ -224,19 +224,20 @@ with gr.Blocks(css='style.css') as demo:
|
|
224 |
output_audio = gr.Audio(label="Edited Audio", interactive=False, scale=1)
|
225 |
|
226 |
with gr.Row():
|
227 |
-
tar_prompt = gr.Textbox(label="Prompt", info="Describe your desired edited output",
|
|
|
228 |
lines=2, interactive=True)
|
229 |
|
230 |
with gr.Row():
|
231 |
t_start = gr.Slider(minimum=15, maximum=85, value=45, step=1, label="T-start (%)", interactive=True, scale=3,
|
232 |
-
info="
|
233 |
-
# model_id = gr.
|
234 |
-
model_id = gr.
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
|
241 |
with gr.Row():
|
242 |
with gr.Column():
|
@@ -244,7 +245,8 @@ with gr.Blocks(css='style.css') as demo:
|
|
244 |
|
245 |
with gr.Accordion("More Options", open=False):
|
246 |
with gr.Row():
|
247 |
-
src_prompt = gr.Textbox(label="Source Prompt", lines=2, interactive=True,
|
|
|
248 |
placeholder="A recording of a happy upbeat classical music piece",)
|
249 |
|
250 |
with gr.Row():
|
@@ -253,6 +255,7 @@ with gr.Blocks(css='style.css') as demo:
|
|
253 |
cfg_scale_tar = gr.Number(value=12, minimum=0.5, maximum=25, precision=None,
|
254 |
label="Target Guidance Scale", interactive=True, scale=1)
|
255 |
steps = gr.Number(value=50, step=1, minimum=20, maximum=300,
|
|
|
256 |
label="Num Diffusion Steps", interactive=True, scale=1)
|
257 |
with gr.Row():
|
258 |
seed = gr.Number(value=0, precision=0, label="Seed", interactive=True)
|
|
|
224 |
output_audio = gr.Audio(label="Edited Audio", interactive=False, scale=1)
|
225 |
|
226 |
with gr.Row():
|
227 |
+
tar_prompt = gr.Textbox(label="Prompt", info="Describe your desired edited output",
|
228 |
+
placeholder="a recording of a happy upbeat arcade game soundtrack",
|
229 |
lines=2, interactive=True)
|
230 |
|
231 |
with gr.Row():
|
232 |
t_start = gr.Slider(minimum=15, maximum=85, value=45, step=1, label="T-start (%)", interactive=True, scale=3,
|
233 |
+
info="Lower T-start -> closer to original audio. Higher T-start -> stronger edit.")
|
234 |
+
# model_id = gr.Radio(label="AudioLDM2 Version",
|
235 |
+
model_id = gr.Dropdown(label="AudioLDM2 Version",
|
236 |
+
choices=["cvssp/audioldm2",
|
237 |
+
"cvssp/audioldm2-large",
|
238 |
+
"cvssp/audioldm2-music"],
|
239 |
+
info="Choose a checkpoint suitable for your intended audio and edit",
|
240 |
+
value="cvssp/audioldm2-music", interactive=True, type="value", scale=2)
|
241 |
|
242 |
with gr.Row():
|
243 |
with gr.Column():
|
|
|
245 |
|
246 |
with gr.Accordion("More Options", open=False):
|
247 |
with gr.Row():
|
248 |
+
src_prompt = gr.Textbox(label="Source Prompt", lines=2, interactive=True,
|
249 |
+
info="Optional: Describe the original audio input",
|
250 |
placeholder="A recording of a happy upbeat classical music piece",)
|
251 |
|
252 |
with gr.Row():
|
|
|
255 |
cfg_scale_tar = gr.Number(value=12, minimum=0.5, maximum=25, precision=None,
|
256 |
label="Target Guidance Scale", interactive=True, scale=1)
|
257 |
steps = gr.Number(value=50, step=1, minimum=20, maximum=300,
|
258 |
+
info="Higher values (e.g. 200) yield higher-quality generation.",
|
259 |
label="Num Diffusion Steps", interactive=True, scale=1)
|
260 |
with gr.Row():
|
261 |
seed = gr.Number(value=0, precision=0, label="Seed", interactive=True)
|