Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -4,60 +4,72 @@ import requests, os, time
|
|
4 |
ASR_API_URL = os.getenv("ASR_API_URL")
|
5 |
AUTH_TOKEN = os.getenv("AUTH_TOKEN")
|
6 |
|
7 |
-
def transcribe_audio(file_path):
|
|
|
8 |
if not file_path:
|
9 |
-
return "β Audio
|
10 |
if not ASR_API_URL or not AUTH_TOKEN:
|
11 |
return "β Error: ASR_API_URL or AUTH_TOKEN is not set.", ""
|
12 |
-
|
13 |
-
headers = {
|
14 |
-
|
15 |
-
|
|
|
|
|
16 |
start = time.time()
|
17 |
try:
|
18 |
-
|
|
|
|
|
19 |
except Exception as e:
|
20 |
return f"β Error: {e}", ""
|
21 |
-
|
22 |
-
|
23 |
if response.status_code == 200:
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
return
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
#gooya-title {color:white; background:linear-gradient(90deg,#224CA5 0%,#2CD8D5 100%);
|
33 |
border-radius:12px;padding:20px 10px;margin-bottom:12px;}
|
34 |
.gooya-badge {display:inline-block;background:#224CA5;color:#fff;border-radius:16px;
|
35 |
padding:6px 16px;font-size:0.97rem;margin-top:4px;}
|
36 |
#gooya-box {background:#F7FAFF;border:1px solid #e7e9ef;border-radius:14px;
|
37 |
padding:22px 18px;margin-top:12px;}
|
38 |
-
"""
|
|
|
|
|
39 |
|
40 |
-
gr.HTML("""
|
|
|
41 |
<h1 style='margin-bottom:10px;font-weight:800;font-size:2rem;'>Gooya ASR
|
42 |
-
<span style="font-size:1.1rem;
|
43 |
<p style='font-size:1.12rem;margin-bottom:2px;'>High-performance Persian Speech-to-Text</p>
|
44 |
-
<p style='font-size:0.98rem;color:#c6e8fa'>
|
45 |
-
|
46 |
-
|
47 |
|
48 |
with gr.Row():
|
49 |
with gr.Column():
|
50 |
-
audio = gr.Audio(
|
51 |
-
|
52 |
-
|
53 |
-
|
|
|
|
|
54 |
with gr.Column():
|
55 |
inf_time_lbl = gr.Label(label="β±οΈ Processing Time", elem_classes="gooya-badge")
|
56 |
-
transcription = gr.Textbox(
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
|
|
|
|
61 |
|
62 |
with gr.Row():
|
63 |
submit_btn = gr.Button("Transcribe", variant="primary", interactive=False)
|
@@ -74,22 +86,28 @@ For benchmarks visit: [Persian ASR Leaderboard](https://huggingface.co/spaces/na
|
|
74 |
def toggle_btn(path):
|
75 |
return gr.Button.update(interactive=bool(path))
|
76 |
|
77 |
-
audio.change(toggle_btn,
|
|
|
78 |
|
79 |
submit_btn.click(
|
80 |
transcribe_audio,
|
81 |
inputs=audio,
|
82 |
-
outputs=[transcription, inf_time_lbl]
|
83 |
).then(
|
84 |
lambda: gr.Button.update(interactive=False),
|
85 |
None,
|
86 |
-
submit_btn
|
|
|
87 |
)
|
88 |
|
|
|
|
|
|
|
89 |
clear_btn.click(
|
90 |
-
|
91 |
None,
|
92 |
[transcription, inf_time_lbl, audio, submit_btn],
|
|
|
93 |
)
|
94 |
|
95 |
demo.launch()
|
|
|
4 |
ASR_API_URL = os.getenv("ASR_API_URL")
|
5 |
AUTH_TOKEN = os.getenv("AUTH_TOKEN")
|
6 |
|
7 |
+
def transcribe_audio(file_path: str | None):
|
8 |
+
# ΪΪ© ΩΨ§ΫΩ
|
9 |
if not file_path:
|
10 |
+
return "β Audio File loading ...", ""
|
11 |
if not ASR_API_URL or not AUTH_TOKEN:
|
12 |
return "β Error: ASR_API_URL or AUTH_TOKEN is not set.", ""
|
13 |
+
|
14 |
+
headers = {
|
15 |
+
"accept": "application/json",
|
16 |
+
"Authorization": f"Bearer {AUTH_TOKEN}",
|
17 |
+
}
|
18 |
+
|
19 |
start = time.time()
|
20 |
try:
|
21 |
+
with open(file_path, "rb") as f:
|
22 |
+
files = {"file": (os.path.basename(file_path), f, "audio/mpeg")}
|
23 |
+
response = requests.post(ASR_API_URL, headers=headers, files=files)
|
24 |
except Exception as e:
|
25 |
return f"β Error: {e}", ""
|
26 |
+
|
27 |
+
elapsed = time.time() - start
|
28 |
if response.status_code == 200:
|
29 |
+
data = response.json()
|
30 |
+
txt = data.get("transcription", "No transcription returned.")
|
31 |
+
elapsed = data.get("time", elapsed)
|
32 |
+
return txt, f"{elapsed:.2f} seconds"
|
33 |
+
return f"β Error {response.status_code}: {response.text}", ""
|
34 |
+
|
35 |
+
|
36 |
+
CUSTOM_CSS = """
|
37 |
#gooya-title {color:white; background:linear-gradient(90deg,#224CA5 0%,#2CD8D5 100%);
|
38 |
border-radius:12px;padding:20px 10px;margin-bottom:12px;}
|
39 |
.gooya-badge {display:inline-block;background:#224CA5;color:#fff;border-radius:16px;
|
40 |
padding:6px 16px;font-size:0.97rem;margin-top:4px;}
|
41 |
#gooya-box {background:#F7FAFF;border:1px solid #e7e9ef;border-radius:14px;
|
42 |
padding:22px 18px;margin-top:12px;}
|
43 |
+
"""
|
44 |
+
|
45 |
+
with gr.Blocks(css=CUSTOM_CSS) as demo:
|
46 |
|
47 |
+
gr.HTML("""
|
48 |
+
<div id="gooya-title">
|
49 |
<h1 style='margin-bottom:10px;font-weight:800;font-size:2rem;'>Gooya ASR
|
50 |
+
<span style="font-size:1.1rem;font-weight:400;opacity:0.8;'>v1.4</span></h1>
|
51 |
<p style='font-size:1.12rem;margin-bottom:2px;'>High-performance Persian Speech-to-Text</p>
|
52 |
+
<p style='font-size:0.98rem;color:#c6e8fa'>Upload or record a Persian audio file (max 30s) and instantly receive the transcription.</p>
|
53 |
+
</div>
|
54 |
+
""")
|
55 |
|
56 |
with gr.Row():
|
57 |
with gr.Column():
|
58 |
+
audio = gr.Audio(
|
59 |
+
label="Audio Input (Upload or record, up to 30s)",
|
60 |
+
type="filepath",
|
61 |
+
sources=["upload", "microphone"],
|
62 |
+
show_label=True,
|
63 |
+
)
|
64 |
with gr.Column():
|
65 |
inf_time_lbl = gr.Label(label="β±οΈ Processing Time", elem_classes="gooya-badge")
|
66 |
+
transcription = gr.Textbox(
|
67 |
+
label="π Transcription",
|
68 |
+
lines=5,
|
69 |
+
show_copy_button=True,
|
70 |
+
placeholder="The transcription will appear here...",
|
71 |
+
elem_id="gooya-textbox",
|
72 |
+
)
|
73 |
|
74 |
with gr.Row():
|
75 |
submit_btn = gr.Button("Transcribe", variant="primary", interactive=False)
|
|
|
86 |
def toggle_btn(path):
|
87 |
return gr.Button.update(interactive=bool(path))
|
88 |
|
89 |
+
audio.change(toggle_btn, inputs=audio, outputs=submit_btn, queue=False)
|
90 |
+
audio.stop_record(toggle_btn, inputs=audio, outputs=submit_btn, queue=False)
|
91 |
|
92 |
submit_btn.click(
|
93 |
transcribe_audio,
|
94 |
inputs=audio,
|
95 |
+
outputs=[transcription, inf_time_lbl],
|
96 |
).then(
|
97 |
lambda: gr.Button.update(interactive=False),
|
98 |
None,
|
99 |
+
submit_btn,
|
100 |
+
queue=False,
|
101 |
)
|
102 |
|
103 |
+
def clear_all():
|
104 |
+
return "", "", None, gr.Button.update(interactive=False)
|
105 |
+
|
106 |
clear_btn.click(
|
107 |
+
clear_all,
|
108 |
None,
|
109 |
[transcription, inf_time_lbl, audio, submit_btn],
|
110 |
+
queue=False,
|
111 |
)
|
112 |
|
113 |
demo.launch()
|