Raivis Dejus
commited on
Commit
β’
adb4549
1
Parent(s):
00936c5
Adjusting notes
Browse files
app.py
CHANGED
@@ -107,21 +107,16 @@ transcribe = gr.Interface(
|
|
107 |
gr.Radio([("Transcribe", "transcribe"), ("Translate to English", "translate",)], label="Task", value="transcribe"),
|
108 |
],
|
109 |
outputs=gr.Textbox(label="Transcription", lines=15),
|
110 |
-
title="Latvian speech recognition:
|
111 |
-
description=("""
|
112 |
-
|
113 |
-
|
114 |
-
<h2>tiny</h2>
|
115 |
-
|
116 |
-
[RaivisDejus/whisper-tiny-lv](https://huggingface.co/RaivisDejus/whisper-tiny-lv) - Fastest, requiring least RAM, but also poor accuracy. On this demo hardware 30 second audio will take ~45 seconds to transcribe.
|
117 |
|
118 |
-
|
|
|
119 |
|
120 |
-
[
|
121 |
-
|
122 |
-
<h2>large</h2>
|
123 |
-
|
124 |
-
[AiLab-IMCS-UL/whisper-large-v3-lv-late-cv17](https://huggingface.co/AiLab-IMCS-UL/whisper-large-v3-lv-late-cv17) - Most accurate, developed by scientists from [ailab.lv](https://ailab.lv/). Requires most RAM and for best performance should be run on a GPU. On this demo hardware 30 second audio will take ~4 minutes to transcribe.
|
125 |
|
126 |
To improve speech recognition quality, more data is needed, add your voice on [Balsu talka](https://balsutalka.lv/)
|
127 |
"""
|
@@ -141,13 +136,11 @@ yt_transcribe = gr.Interface(
|
|
141 |
],
|
142 |
# outputs=["html", "text"],
|
143 |
outputs=[gr.HTML(), gr.Textbox(label="Transcription", lines=10)],
|
144 |
-
title="Latvian speech recognition:
|
145 |
description=("""
|
146 |
-
|
147 |
-
|
148 |
-
* [tiny](https://huggingface.co/RaivisDejus/whisper-tiny-lv) - Fastest, requiring least RAM, but also poor accuracy
|
149 |
|
150 |
-
|
151 |
|
152 |
To improve speech recognition quality, more data is needed, add your voice on [Balsu talka](https://balsutalka.lv/)
|
153 |
"""
|
|
|
107 |
gr.Radio([("Transcribe", "transcribe"), ("Translate to English", "translate",)], label="Task", value="transcribe"),
|
108 |
],
|
109 |
outputs=gr.Textbox(label="Transcription", lines=15),
|
110 |
+
title="Latvian speech recognition: Three models available",
|
111 |
+
description=("""
|
112 |
+
π€ [tiny](https://huggingface.co/RaivisDejus/whisper-tiny-lv) - Fastest, requiring least RAM, but also poor accuracy.
|
113 |
+
On this demo hardware 30 second audio will take ~45 seconds to transcribe.
|
|
|
|
|
|
|
114 |
|
115 |
+
π€ [small](https://huggingface.co/RaivisDejus/whisper-small-lv) - Reasonably fast, reasonably accurate, requiring reasonable amounts of RAM.
|
116 |
+
On this demo hardware 30 second audio will take ~1 minute to transcribe.
|
117 |
|
118 |
+
π€ [large](https://huggingface.co/AiLab-IMCS-UL/whisper-large-v3-lv-late-cv17) - Most accurate, developed by scientists from [ailab.lv](https://ailab.lv/). Requires most RAM and for best performance should be run on a GPU.
|
119 |
+
On this demo hardware 30 second audio will take ~4 minutes to transcribe.
|
|
|
|
|
|
|
120 |
|
121 |
To improve speech recognition quality, more data is needed, add your voice on [Balsu talka](https://balsutalka.lv/)
|
122 |
"""
|
|
|
136 |
],
|
137 |
# outputs=["html", "text"],
|
138 |
outputs=[gr.HTML(), gr.Textbox(label="Transcription", lines=10)],
|
139 |
+
title="Latvian speech recognition: Two models available",
|
140 |
description=("""
|
141 |
+
π€ [tiny](https://huggingface.co/RaivisDejus/whisper-tiny-lv) - Fastest, requiring least RAM, but also poor accuracy
|
|
|
|
|
142 |
|
143 |
+
π€ [small](https://huggingface.co/RaivisDejus/whisper-small-lv) - Reasonably fast, reasonably accurate, requiring reasonable amounts of RAM
|
144 |
|
145 |
To improve speech recognition quality, more data is needed, add your voice on [Balsu talka](https://balsutalka.lv/)
|
146 |
"""
|