Added multilingual caption support

speech-to-text/README.md

@@ -6,23 +6,24 @@ Convert audio to text using Smallest AI's Pulse Speech-to-Text API. Supports 30+
 
 ## Examples
 
-| Example | Description |
-|---------|-------------|
-| [Getting Started](./getting-started/) | Basic transcription - the simplest way to get started |
-| [Word-Level Outputs](./word-level-outputs/) | Word timestamps and speaker diarization |
-| [Subtitle Generation](./subtitle-generation/) | Generate SRT/VTT subtitles from audio or video |
-| [Meeting Notes](./meeting-notes/) | Join meetings via Recall.ai, auto-identify speakers by name |
-| [Podcast Summarizer](./podcast-summarizer/) | Transcribe and summarize podcasts with GPT-4o |
-| [File Transcription](./file-transcription/) | Transcribe files with all advanced features |
-| [Emotion Analyzer](./emotion-analyzer/) | Visualize speaker emotions across a conversation with interactive charts |
+| Example                                                               | Description                                                              |
+| --------------------------------------------------------------------- | ------------------------------------------------------------------------ |
+| [Getting Started](./getting-started/)                                 | Basic transcription - the simplest way to get started                    |
+| [Word-Level Outputs](./word-level-outputs/)                           | Word timestamps and speaker diarization                                  |
+| [Subtitle Generation](./subtitle-generation/)                         | Generate SRT/VTT subtitles from audio or video                           |
+| [Multilingual Live Captions](./websocket/multilingual-live-captions/) | Auto-detect language, translate, and preview live SRT captions           |
+| [Meeting Notes](./meeting-notes/)                                     | Join meetings via Recall.ai, auto-identify speakers by name              |
+| [Podcast Summarizer](./podcast-summarizer/)                           | Transcribe and summarize podcasts with GPT-4o                            |
+| [File Transcription](./file-transcription/)                           | Transcribe files with all advanced features                              |
+| [Emotion Analyzer](./emotion-analyzer/)                               | Visualize speaker emotions across a conversation with interactive charts |
 
 ### WebSocket Examples
 
-| Example | Description |
-|---------|-------------|
-| [Streaming Transcription](./websocket/streaming-text-output-transcription/) | Stream audio files via WebSocket |
-| [Realtime Microphone](./websocket/realtime-microphone-transcription/) | Gradio web UI with live microphone transcription |
-| [Jarvis Voice Assistant](./websocket/jarvis/) | Always-on assistant with wake word, LLM, and TTS |
+| Example                                                                     | Description                                      |
+| --------------------------------------------------------------------------- | ------------------------------------------------ |
+| [Streaming Transcription](./websocket/streaming-text-output-transcription/) | Stream audio files via WebSocket                 |
+| [Realtime Microphone](./websocket/realtime-microphone-transcription/)       | Gradio web UI with live microphone transcription |
+| [Jarvis Voice Assistant](./websocket/jarvis/)                               | Always-on assistant with wake word, LLM, and TTS |
 
 ## Quick Start
 

speech-to-text/subtitle-generation/python/transcribe.py

@@ -19,9 +19,12 @@
 import requests
 from dotenv import load_dotenv
 
+sys.path.insert(0, str(Path(__file__).parent.parent.parent))
+from subtitle_utils import format_timestamp_srt, format_timestamp_vtt, generate_srt, generate_vtt
+
 load_dotenv()
 
-API_URL = "https://waves-api.smallest.ai/api/v1/pulse/get_text"
+API_URL = "https://api.smallest.ai/waves/v1/pulse/get_text"
 
 LANGUAGE = "en"  # Use ISO 639-1 codes or "multi" for auto-detect
 WORDS_PER_SEGMENT = 10  # Maximum words per subtitle segment
@@ -52,24 +55,6 @@ def transcribe(audio_file: str, api_key: str) -> dict:
     return response.json()
 
 
-def format_time_srt(seconds: float) -> str:
-    """Format seconds to SRT timestamp: HH:MM:SS,mmm"""
-    hours = int(seconds // 3600)
-    minutes = int((seconds % 3600) // 60)
-    secs = int(seconds % 60)
-    millis = int((seconds % 1) * 1000)
-    return f"{hours:02d}:{minutes:02d}:{secs:02d},{millis:03d}"
-
-
-def format_time_vtt(seconds: float) -> str:
-    """Format seconds to VTT timestamp: HH:MM:SS.mmm"""
-    hours = int(seconds // 3600)
-    minutes = int((seconds % 3600) // 60)
-    secs = int(seconds % 60)
-    millis = int((seconds % 1) * 1000)
-    return f"{hours:02d}:{minutes:02d}:{secs:02d}.{millis:03d}"
-
-
 def create_segments(words: list) -> list:
     """Group words into subtitle segments."""
     segments = []
@@ -103,32 +88,6 @@ def create_segments(words: list) -> list:
     return segments
 
 
-def generate_srt(segments: list) -> str:
-    """Generate SRT format subtitles."""
-    lines = []
-    for i, segment in enumerate(segments, 1):
-        start = format_time_srt(segment["start"])
-        end = format_time_srt(segment["end"])
-        lines.append(f"{i}")
-        lines.append(f"{start} --> {end}")
-        lines.append(segment["text"])
-        lines.append("")
-    return "\n".join(lines)
-
-
-def generate_vtt(segments: list) -> str:
-    """Generate WebVTT format subtitles."""
-    lines = ["WEBVTT", ""]
-    for i, segment in enumerate(segments, 1):
-        start = format_time_vtt(segment["start"])
-        end = format_time_vtt(segment["end"])
-        lines.append(f"{i}")
-        lines.append(f"{start} --> {end}")
-        lines.append(segment["text"])
-        lines.append("")
-    return "\n".join(lines)
-
-
 def process_response(result: dict, audio_path: Path):
     if result.get("status") != "success":
         print("Error: Transcription failed")

speech-to-text/subtitle_utils.py

@@ -0,0 +1,152 @@
+"""
+Shared subtitle utilities for Smallest AI Speech-to-Text applications.
+
+This module provides reusable functions for generating SRT and VTT subtitle
+formats with proper timestamp formatting.
+"""
+
+from typing import List, Dict, Callable, Optional
+
+
+def format_timestamp_srt(seconds: float) -> str:
+    """
+    Format seconds to SRT timestamp format: HH:MM:SS,mmm
+
+    Args:
+        seconds: Time in seconds (can be float)
+
+    Returns:
+        Formatted timestamp string in SRT format
+
+    Example:
+        >>> format_timestamp_srt(65.123)
+        '00:01:05,123'
+    """
+    ms = int(seconds * 1000)
+    hours, ms = divmod(ms, 3600 * 1000)
+    minutes, ms = divmod(ms, 60 * 1000)
+    secs, ms = divmod(ms, 1000)
+    return f"{hours:02d}:{minutes:02d}:{secs:02d},{ms:03d}"
+
+
+def format_timestamp_vtt(seconds: float) -> str:
+    """
+    Format seconds to VTT timestamp format: HH:MM:SS.mmm
+
+    Args:
+        seconds: Time in seconds (can be float)
+
+    Returns:
+        Formatted timestamp string in VTT format
+
+    Example:
+        >>> format_timestamp_vtt(65.123)
+        '00:01:05.123'
+    """
+    ms = int(seconds * 1000)
+    hours, ms = divmod(ms, 3600 * 1000)
+    minutes, ms = divmod(ms, 60 * 1000)
+    secs, ms = divmod(ms, 1000)
+    return f"{hours:02d}:{minutes:02d}:{secs:02d}.{ms:03d}"
+
+
+def generate_srt(
+    entries: List[Dict],
+    text_key: str = "text",
+    start_key: str = "start",
+    end_key: str = "end",
+    format_func: Optional[Callable[[float], str]] = None
+) -> str:
+    """
+    Generate SRT format subtitles from a list of entries.
+
+    Args:
+        entries: List of dictionaries containing subtitle data
+        text_key: Key name for the text content in each entry
+        start_key: Key name for the start timestamp in each entry
+        end_key: Key name for the end timestamp in each entry
+        format_func: Optional custom timestamp formatting function.
+                    Defaults to format_timestamp_srt.
+
+    Returns:
+        Complete SRT subtitle string
+
+    Example:
+        >>> entries = [
+        ...     {"text": "Hello", "start": 0.0, "end": 1.5},
+        ...     {"text": "World", "start": 1.5, "end": 3.0}
+        ... ]
+        >>> print(generate_srt(entries))
+        1
+        00:00:00,000 --> 00:00:01,500
+        Hello
+
+        2
+        00:00:01,500 --> 00:00:03,000
+        World
+    """
+    if format_func is None:
+        format_func = format_timestamp_srt
+
+    lines = []
+    for idx, entry in enumerate(entries, start=1):
+        start = format_func(entry[start_key])
+        end = format_func(entry[end_key])
+        lines.append(str(idx))
+        lines.append(f"{start} --> {end}")
+        lines.append(entry[text_key])
+        lines.append("")
+
+    return "\n".join(lines).strip()
+
+
+def generate_vtt(
+    entries: List[Dict],
+    text_key: str = "text",
+    start_key: str = "start",
+    end_key: str = "end",
+    format_func: Optional[Callable[[float], str]] = None
+) -> str:
+    """
+    Generate WebVTT format subtitles from a list of entries.
+
+    Args:
+        entries: List of dictionaries containing subtitle data
+        text_key: Key name for the text content in each entry
+        start_key: Key name for the start timestamp in each entry
+        end_key: Key name for the end timestamp in each entry
+        format_func: Optional custom timestamp formatting function.
+                    Defaults to format_timestamp_vtt.
+
+    Returns:
+        Complete WebVTT subtitle string
+
+    Example:
+        >>> entries = [
+        ...     {"text": "Hello", "start": 0.0, "end": 1.5},
+        ...     {"text": "World", "start": 1.5, "end": 3.0}
+        ... ]
+        >>> print(generate_vtt(entries))
+        WEBVTT
+
+        1
+        00:00:00.000 --> 00:00:01.500
+        Hello
+
+        2
+        00:00:01.500 --> 00:00:03.000
+        World
+    """
+    if format_func is None:
+        format_func = format_timestamp_vtt
+
+    lines = ["WEBVTT", ""]
+    for idx, entry in enumerate(entries, start=1):
+        start = format_func(entry[start_key])
+        end = format_func(entry[end_key])
+        lines.append(str(idx))
+        lines.append(f"{start} --> {end}")
+        lines.append(entry[text_key])
+        lines.append("")
+
+    return "\n".join(lines).strip()

speech-to-text/websocket/multilingual-live-captions/.env.sample

@@ -0,0 +1,5 @@
+# Smallest AI API Key
+SMALLEST_API_KEY=your-smallest-api-key
+
+# Optional: translation (falls back to source if unset)
+OPENAI_API_KEY=your-openai-api-key

speech-to-text/websocket/multilingual-live-captions/README.md

@@ -0,0 +1,78 @@
+# Multilingual Live Captions
+
+Auto-detect spoken language, translate on the fly, and preview SRT captions for live events or streams.
+
+## Features
+
+- Automatic language detection (`language=multi`) with Pulse STT
+- Live translation to a chosen subtitle language
+- Instant SRT preview you can copy into players/overlays
+- Works from your microphone via WebSocket streaming
+
+## Requirements
+
+> Base dependencies are installed via the root `requirements.txt`. See the [main README](../../../README.md#usage) for setup. Add `SMALLEST_API_KEY` to your `.env`.
+
+Extra dependencies:
+
+```bash
+uv pip install -r requirements.txt
+```
+
+This installs `numpy` (used for lightweight resampling). Translation uses OpenAI if `OPENAI_API_KEY` is set; otherwise it falls back to source text.
+
+## Usage
+
+```bash
+uv run app.py
+```
+
+Then open http://localhost:7860 and:
+
+1. Choose subtitle language (e.g., English)
+2. Click microphone and speak — transcript + translation update live
+3. Copy SRT preview into your caption tool or player
+4. Click **Stop & Clear** to reset
+
+## Recommended Usage
+
+- Live captions for multilingual events or webinars
+- Streaming overlays that need translated subtitles quickly
+- As a building block for hybrid ASR + translation caption services
+
+## How It Works
+
+- Streams mic audio to Pulse STT over WebSocket with `language=multi`
+- Receives partial/final transcripts with detected language metadata
+- Runs on-the-fly translation via OpenAI (if configured)
+- Accumulates segments and renders an SRT preview for copy/paste
+
+## Configuration
+
+| Setting           | Default   | Description                                                   |
+| ----------------- | --------- | ------------------------------------------------------------- |
+| `language`        | `multi`   | Enables automatic language detection                          |
+| Subtitle dropdown | `English` | Target translation language (`Same as source` keeps original) |
+| Sample rate       | 16000 Hz  | Resampled client-side if needed                               |
+
+## Example Output (SRT)
+
+```
+1
+00:00:00,000 --> 00:00:03,400
+Welcome to the live demo.
+
+2
+00:00:03,400 --> 00:00:06,200
+Here are your translated captions.
+```
+
+## API Reference
+
+- [Streaming Quickstart](https://waves-docs.smallest.ai/v4.0.0/content/speech-to-text-new/streaming/quickstart)
+- [Pulse STT WebSocket API](https://waves-docs.smallest.ai/content/api-references/pulse-stt-ws)
+
+## Next Steps
+
+- [Realtime Microphone Transcription](../realtime-microphone-transcription/) — Basic live STT without translation
+- [Jarvis Voice Assistant](../jarvis/) — Full assistant with wake word, LLM, and TTS