Spaces:
Sleeping
Sleeping
Commit
Β·
7304960
1
Parent(s):
079cf4d
Debugging
Browse files
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
|
@@ -67,7 +68,7 @@ def get_translation_history(direction):
|
|
67 |
# --- Translation ---
|
68 |
def translate(text, direction):
|
69 |
if not text.strip():
|
70 |
-
return "Enter some text.",
|
71 |
|
72 |
if direction == "en_to_ks":
|
73 |
src_lang, tgt_lang = "eng_Latn", "kas_Arab"
|
@@ -83,12 +84,12 @@ def translate(text, direction):
|
|
83 |
output = model.generate(**tokens, max_length=256, num_beams=5)
|
84 |
result = tokenizer.batch_decode(output, skip_special_tokens=True)
|
85 |
final = ip.postprocess_batch(result, lang=tgt_lang)[0]
|
86 |
-
return final,
|
87 |
except Exception as e:
|
88 |
print("Translation error:", e)
|
89 |
-
return "β οΈ Translation failed.",
|
90 |
|
91 |
-
# --- TTS for English output
|
92 |
def synthesize_tts(text):
|
93 |
try:
|
94 |
tts = gTTS(text=text, lang="en")
|
@@ -100,25 +101,31 @@ def synthesize_tts(text):
|
|
100 |
return None
|
101 |
|
102 |
# --- STT only for en_to_ks ---
|
103 |
-
def
|
104 |
if direction != "en_to_ks":
|
105 |
-
return "β οΈ Audio input is only supported for English to Kashmiri.", "",
|
106 |
|
107 |
try:
|
108 |
transcription = asr(audio_path)["text"]
|
109 |
except Exception as e:
|
110 |
print("STT error:", e)
|
111 |
-
return "β οΈ Transcription failed.", "",
|
112 |
|
113 |
-
translated, _
|
114 |
-
return transcription, translated,
|
|
|
|
|
|
|
|
|
|
|
|
|
115 |
|
116 |
# --- Switch UI direction ---
|
117 |
def switch_direction(direction, input_text_val, output_text_val):
|
118 |
new_direction = "ks_to_en" if direction == "en_to_ks" else "en_to_ks"
|
119 |
input_label = "Kashmiri Text" if new_direction == "ks_to_en" else "English Text"
|
120 |
output_label = "English Translation" if new_direction == "ks_to_en" else "Kashmiri Translation"
|
121 |
-
return new_direction, gr.update(value=output_text_val, label=input_label), gr.update(value=input_text_val, label=output_label)
|
122 |
|
123 |
# === Gradio Interface ===
|
124 |
with gr.Blocks() as interface:
|
@@ -126,8 +133,8 @@ with gr.Blocks() as interface:
|
|
126 |
translation_direction = gr.State(value="en_to_ks")
|
127 |
|
128 |
with gr.Row():
|
129 |
-
input_text = gr.Textbox(label="English Text", placeholder="Enter text here...")
|
130 |
-
output_text = gr.Textbox(label="Kashmiri Translation", placeholder="Translated text...")
|
131 |
|
132 |
with gr.Row():
|
133 |
translate_button = gr.Button("Translate")
|
@@ -139,21 +146,45 @@ with gr.Blocks() as interface:
|
|
139 |
|
140 |
with gr.Row():
|
141 |
audio_input = gr.Audio(type="filepath", label="ποΈ Record English audio")
|
142 |
-
audio_output = gr.Audio(label="π English TTS")
|
143 |
|
144 |
stt_button = gr.Button("π€ Transcribe & Translate (EN β KS Only)")
|
|
|
145 |
|
146 |
# Events
|
147 |
-
translate_button.click(
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
157 |
|
158 |
if __name__ == "__main__":
|
159 |
-
interface.queue().launch()
|
|
|
1 |
+
#app.py
|
2 |
import gradio as gr
|
3 |
import torch
|
4 |
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
|
|
|
68 |
# --- Translation ---
|
69 |
def translate(text, direction):
|
70 |
if not text.strip():
|
71 |
+
return "Enter some text.", None
|
72 |
|
73 |
if direction == "en_to_ks":
|
74 |
src_lang, tgt_lang = "eng_Latn", "kas_Arab"
|
|
|
84 |
output = model.generate(**tokens, max_length=256, num_beams=5)
|
85 |
result = tokenizer.batch_decode(output, skip_special_tokens=True)
|
86 |
final = ip.postprocess_batch(result, lang=tgt_lang)[0]
|
87 |
+
return final, None
|
88 |
except Exception as e:
|
89 |
print("Translation error:", e)
|
90 |
+
return "β οΈ Translation failed.", None
|
91 |
|
92 |
+
# --- TTS for English output ---
|
93 |
def synthesize_tts(text):
|
94 |
try:
|
95 |
tts = gTTS(text=text, lang="en")
|
|
|
101 |
return None
|
102 |
|
103 |
# --- STT only for en_to_ks ---
|
104 |
+
def generate_stt_for_input(audio_path, direction):
|
105 |
if direction != "en_to_ks":
|
106 |
+
return "β οΈ Audio input is only supported for English to Kashmiri.", "", None
|
107 |
|
108 |
try:
|
109 |
transcription = asr(audio_path)["text"]
|
110 |
except Exception as e:
|
111 |
print("STT error:", e)
|
112 |
+
return "β οΈ Transcription failed.", "", None
|
113 |
|
114 |
+
translated, _ = translate(transcription, direction)
|
115 |
+
return transcription, translated, None
|
116 |
+
|
117 |
+
# --- Generate TTS for English output ---
|
118 |
+
def generate_tts_for_output(output_text, direction):
|
119 |
+
if direction == "ks_to_en" and output_text.strip():
|
120 |
+
return synthesize_tts(output_text)
|
121 |
+
return None
|
122 |
|
123 |
# --- Switch UI direction ---
|
124 |
def switch_direction(direction, input_text_val, output_text_val):
|
125 |
new_direction = "ks_to_en" if direction == "en_to_ks" else "en_to_ks"
|
126 |
input_label = "Kashmiri Text" if new_direction == "ks_to_en" else "English Text"
|
127 |
output_label = "English Translation" if new_direction == "ks_to_en" else "Kashmiri Translation"
|
128 |
+
return new_direction, gr.update(value=output_text_val, label=input_label), gr.update(value=input_text_val, label=output_label), None
|
129 |
|
130 |
# === Gradio Interface ===
|
131 |
with gr.Blocks() as interface:
|
|
|
133 |
translation_direction = gr.State(value="en_to_ks")
|
134 |
|
135 |
with gr.Row():
|
136 |
+
input_text = gr.Textbox(label="English Text", placeholder="Enter text here...", lines=2)
|
137 |
+
output_text = gr.Textbox(label="Kashmiri Translation", placeholder="Translated text...", lines=2)
|
138 |
|
139 |
with gr.Row():
|
140 |
translate_button = gr.Button("Translate")
|
|
|
146 |
|
147 |
with gr.Row():
|
148 |
audio_input = gr.Audio(type="filepath", label="ποΈ Record English audio")
|
149 |
+
audio_output = gr.Audio(label="π English TTS", interactive=False)
|
150 |
|
151 |
stt_button = gr.Button("π€ Transcribe & Translate (EN β KS Only)")
|
152 |
+
tts_button = gr.Button("π Generate English Speech (KS β EN Only)")
|
153 |
|
154 |
# Events
|
155 |
+
translate_button.click(
|
156 |
+
fn=translate,
|
157 |
+
inputs=[input_text, translation_direction],
|
158 |
+
outputs=[output_text, audio_output]
|
159 |
+
)
|
160 |
+
|
161 |
+
tts_button.click(
|
162 |
+
fn=generate_tts_for_output,
|
163 |
+
inputs=[output_text, translation_direction],
|
164 |
+
outputs=audio_output
|
165 |
+
)
|
166 |
+
|
167 |
+
save_button.click(
|
168 |
+
fn=save_to_supabase,
|
169 |
+
inputs=[input_text, output_text, translation_direction],
|
170 |
+
outputs=save_status
|
171 |
+
).then(
|
172 |
+
fn=get_translation_history,
|
173 |
+
inputs=translation_direction,
|
174 |
+
outputs=history
|
175 |
+
)
|
176 |
+
|
177 |
+
switch_button.click(
|
178 |
+
fn=switch_direction,
|
179 |
+
inputs=[translation_direction, input_text, output_text],
|
180 |
+
outputs=[translation_direction, input_text, output_text, audio_output]
|
181 |
+
)
|
182 |
+
|
183 |
+
stt_button.click(
|
184 |
+
fn=generate_stt_for_input,
|
185 |
+
inputs=[audio_input, translation_direction],
|
186 |
+
outputs=[input_text, output_text, audio_output]
|
187 |
+
)
|
188 |
|
189 |
if __name__ == "__main__":
|
190 |
+
interface.queue().launch(share=True)
|