harsh2ai
commited on
Commit
Β·
b672ef4
1
Parent(s):
00c3484
Rebrand to Ringg Parrot STT V1
Browse files
README.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
---
|
| 2 |
-
title: Ringg STT
|
| 3 |
-
emoji:
|
| 4 |
colorFrom: blue
|
| 5 |
colorTo: purple
|
| 6 |
sdk: gradio
|
|
@@ -22,7 +22,7 @@ tags:
|
|
| 22 |
- real-time
|
| 23 |
---
|
| 24 |
|
| 25 |
-
# ποΈ Ringg STT
|
| 26 |
|
| 27 |
**Bilingual Speech-to-Text for English & Hindi**
|
| 28 |
|
|
@@ -31,19 +31,19 @@ tags:
|
|
| 31 |
|
| 32 |
## π Overview
|
| 33 |
|
| 34 |
-
Ringg STT
|
| 35 |
|
| 36 |
## π Performance Benchmarks
|
| 37 |
|
| 38 |
| Model | Indic Norm WER β | Whisper Norm WER β |
|
| 39 |
|-------|------------------|---------------------|
|
| 40 |
| IndicWav2Vec (Winner) | 18.55% | 63.31% |
|
| 41 |
-
| **Ringg STT
|
| 42 |
| VakyanSh Wav2Vec2 | 24.06% | 66.34% |
|
| 43 |
| Whisper Large-v3 | 29.17% | 63.31% |
|
| 44 |
| Whisper Large-v2 | 37.50% | 66.27% |
|
| 45 |
|
| 46 |
-
**Lower WER (Word Error Rate) indicates better accuracy.** Ringg STT
|
| 47 |
|
| 48 |
## β¨ Features
|
| 49 |
|
|
@@ -58,7 +58,7 @@ Ringg STT V0 is a state-of-the-art speech-to-text system that provides real-time
|
|
| 58 |
|
| 59 |
| Specification | Details |
|
| 60 |
|--------------|---------|
|
| 61 |
-
| **Model Name** | Ringg STT
|
| 62 |
| **Languages** | English (EN) & Hindi (HI) |
|
| 63 |
| **Performance** | 2nd place among top models |
|
| 64 |
| **Sample Rate** | 16kHz |
|
|
|
|
| 1 |
---
|
| 2 |
+
title: Ringg Parrot STT V1
|
| 3 |
+
emoji: π¦
|
| 4 |
colorFrom: blue
|
| 5 |
colorTo: purple
|
| 6 |
sdk: gradio
|
|
|
|
| 22 |
- real-time
|
| 23 |
---
|
| 24 |
|
| 25 |
+
# ποΈ Ringg Parrot STT V1 :parrot:
|
| 26 |
|
| 27 |
**Bilingual Speech-to-Text for English & Hindi**
|
| 28 |
|
|
|
|
| 31 |
|
| 32 |
## π Overview
|
| 33 |
|
| 34 |
+
Ringg Parrot STT V1 is a state-of-the-art speech-to-text system that provides real-time transcription for English and Hindi languages. Our model ranks **1st place** among top bilingual ASR models, outperforming OpenAI Whisper Large-v3 and other leading solutions.
|
| 35 |
|
| 36 |
## π Performance Benchmarks
|
| 37 |
|
| 38 |
| Model | Indic Norm WER β | Whisper Norm WER β |
|
| 39 |
|-------|------------------|---------------------|
|
| 40 |
| IndicWav2Vec (Winner) | 18.55% | 63.31% |
|
| 41 |
+
| **Ringg Parrot STT V1** | **21.03%** | **66.27%** |
|
| 42 |
| VakyanSh Wav2Vec2 | 24.06% | 66.34% |
|
| 43 |
| Whisper Large-v3 | 29.17% | 63.31% |
|
| 44 |
| Whisper Large-v2 | 37.50% | 66.27% |
|
| 45 |
|
| 46 |
+
**Lower WER (Word Error Rate) indicates better accuracy.** Ringg Parrot STT V1 achieves competitive performance while supporting bilingual transcription.
|
| 47 |
|
| 48 |
## β¨ Features
|
| 49 |
|
|
|
|
| 58 |
|
| 59 |
| Specification | Details |
|
| 60 |
|--------------|---------|
|
| 61 |
+
| **Model Name** | Ringg Parrot STT V1 |
|
| 62 |
| **Languages** | English (EN) & Hindi (HI) |
|
| 63 |
| **Performance** | 2nd place among top models |
|
| 64 |
| **Sample Rate** | 16kHz |
|
app.py
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
#!/usr/bin/env python3
|
| 2 |
#updated
|
| 3 |
"""
|
| 4 |
-
Ringg STT
|
| 5 |
Makes API calls to private inference endpoint via ngrok
|
| 6 |
"""
|
| 7 |
|
|
@@ -146,7 +146,7 @@ API_ENDPOINT = os.environ.get("STT_API_ENDPOINT", "")
|
|
| 146 |
|
| 147 |
|
| 148 |
class RinggSTTClient:
|
| 149 |
-
"""Client for Ringg STT API"""
|
| 150 |
|
| 151 |
def __init__(self, api_endpoint: str):
|
| 152 |
self.api_endpoint = api_endpoint.rstrip("/")
|
|
@@ -234,7 +234,7 @@ def create_interface():
|
|
| 234 |
|
| 235 |
# Create interface
|
| 236 |
with gr.Blocks(
|
| 237 |
-
title="Ringg STT
|
| 238 |
) as demo:
|
| 239 |
status_class = "healthy" if health_status["status"] == "healthy" else "error"
|
| 240 |
if LOGO_URL:
|
|
@@ -250,7 +250,7 @@ def create_interface():
|
|
| 250 |
<div class="main-header">
|
| 251 |
{logo_html}
|
| 252 |
<div class="main-text">
|
| 253 |
-
<h1>Ringg STT
|
| 254 |
<p>High-Accuracy Hindi Speech-to-Text <span class="status-dot {status_class}"></span></p>
|
| 255 |
</div>
|
| 256 |
</div>
|
|
@@ -259,14 +259,14 @@ def create_interface():
|
|
| 259 |
gr.Markdown(
|
| 260 |
"""
|
| 261 |
# π― Performance Benchmarks
|
| 262 |
-
#### **Ringg STT
|
| 263 |
"""
|
| 264 |
)
|
| 265 |
|
| 266 |
with gr.Row():
|
| 267 |
gr.DataFrame(
|
| 268 |
value=[
|
| 269 |
-
["
|
| 270 |
["IndicWav2Vec ", "19.35%", "20.91%"],
|
| 271 |
["VakyanSh Wav2Vec2", "22.73%", "24.78%"],
|
| 272 |
],
|
|
@@ -345,7 +345,7 @@ def create_interface():
|
|
| 345 |
|
| 346 |
# Launch the app
|
| 347 |
if __name__ == "__main__":
|
| 348 |
-
print("π Launching Ringg STT
|
| 349 |
demo = create_interface()
|
| 350 |
demo.queue(default_concurrency_limit=2, max_size=20)
|
| 351 |
demo.launch(
|
|
|
|
| 1 |
#!/usr/bin/env python3
|
| 2 |
#updated
|
| 3 |
"""
|
| 4 |
+
Ringg Parrot STT V1 π¦ - Hugging Face Space (Frontend)
|
| 5 |
Makes API calls to private inference endpoint via ngrok
|
| 6 |
"""
|
| 7 |
|
|
|
|
| 146 |
|
| 147 |
|
| 148 |
class RinggSTTClient:
|
| 149 |
+
"""Client for Ringg Parrot STT API"""
|
| 150 |
|
| 151 |
def __init__(self, api_endpoint: str):
|
| 152 |
self.api_endpoint = api_endpoint.rstrip("/")
|
|
|
|
| 234 |
|
| 235 |
# Create interface
|
| 236 |
with gr.Blocks(
|
| 237 |
+
title="Ringg Parrot STT V1 π¦", theme=gr.themes.Soft(), css=custom_css
|
| 238 |
) as demo:
|
| 239 |
status_class = "healthy" if health_status["status"] == "healthy" else "error"
|
| 240 |
if LOGO_URL:
|
|
|
|
| 250 |
<div class="main-header">
|
| 251 |
{logo_html}
|
| 252 |
<div class="main-text">
|
| 253 |
+
<h1>Ringg Parrot STT V1 π¦</h1>
|
| 254 |
<p>High-Accuracy Hindi Speech-to-Text <span class="status-dot {status_class}"></span></p>
|
| 255 |
</div>
|
| 256 |
</div>
|
|
|
|
| 259 |
gr.Markdown(
|
| 260 |
"""
|
| 261 |
# π― Performance Benchmarks
|
| 262 |
+
#### **Ringg Parrot STT V1** Ranks **1st** Among Top Models, Outperforming OpenAI Whisper Large-v3 and Other Leading Solutions.
|
| 263 |
"""
|
| 264 |
)
|
| 265 |
|
| 266 |
with gr.Row():
|
| 267 |
gr.DataFrame(
|
| 268 |
value=[
|
| 269 |
+
["Parrot STT (Ringg AI)", "15.00%", "15.92%"],
|
| 270 |
["IndicWav2Vec ", "19.35%", "20.91%"],
|
| 271 |
["VakyanSh Wav2Vec2", "22.73%", "24.78%"],
|
| 272 |
],
|
|
|
|
| 345 |
|
| 346 |
# Launch the app
|
| 347 |
if __name__ == "__main__":
|
| 348 |
+
print("π Launching Ringg Parrot STT V1 Gradio Interface...")
|
| 349 |
demo = create_interface()
|
| 350 |
demo.queue(default_concurrency_limit=2, max_size=20)
|
| 351 |
demo.launch(
|