Audio Response

[Topic] [Duration] [Key Point]

Examples:

• [Disk Space] [15s] EBS volume solved capacity crisis
• [Transcription] [8s] Whisper now working with 95% accuracy
• [TTS Setup] [12s] espeak configured for basic audio responses

TTS Options

1. Local (espeak) - Available now

echo "Response text" | espeak --stdout > response.wav

• Pros: No API, works offline, fast
• Cons: Robotic voice, basic quality

2. OpenAI TTS - Needs API key

from openai import OpenAI
client = OpenAI()
response = client.audio.speech.create(
    model="tts-1",
    voice="alloy",
    input="Response text"
)

3. ElevenLabs - Best quality

from elevenlabs import generate, play
audio = generate(text="Response text", voice="Rachel")

4. Google TTS - Free, requires internet

from gtts import gTTS
tts = gTTS(text="Response text", lang="en")
tts.save("response.mp3")

Implementation Scripts

scripts/audio_responder.py

Main script that:

1. Receives audio file path
2. Transcribes using Whisper
3. Generates response
4. Converts to audio
5. Sends with label

scripts/tts_engine.py

TTS conversion with fallback:

1. Try ElevenLabs if configured
2. Fall back to OpenAI TTS
3. Fall back to Google TTS
4. Final fallback: espeak

scripts/label_generator.py

Generate useful labels:

• Extract topic from conversation
• Estimate duration
• Identify key point

Workflow

User Audio → Transcription → Processing → Text Response → TTS → Labeled Audio Response

Quick Start

Using espeak (immediate):

python3 scripts/audio_responder.py \
  --audio /path/to/user_audio.ogg \
  --tts-engine espeak \
  --label-format "[{topic}] [{duration}s] {key_point}"

Testing:

# Test TTS
echo "Audio response protocol is now active" | espeak --stdout > test.wav

# Test full workflow
python3 scripts/audio_responder.py --test

Integration with OpenClaw

Telegram Handler:

def handle_telegram_audio(audio_path):
    # Transcribe
    transcript = transcribe_audio(audio_path)
    
    # Generate response
    response_text = generate_response(transcript)
    
    # Convert to audio
    audio_response = text_to_speech(response_text)
    
    # Send with label
    label = generate_label(transcript, response_text)
    send_audio_response(audio_response, label)

Auto-Configuration:

Script checks and configures best available TTS:

1. Check for ElevenLabs API key
2. Check for OpenAI API key
3. Check for Google TTS (internet)
4. Default to espeak

Label Generation Examples

Input: "How did you fix the disk space issue?" Output Label: [Infrastructure] [18s] Added 50GB EBS volume, now 47GB free

Input: "Can you transcribe audio now?" Output Label: [Capabilities] [10s] Whisper transcription working with 9 files processed

Input: "What's next?" Output Label: [Planning] [14s] Fix Moltbook cron, set up auto-transcription

Error Handling

TTS Failure:

• Fall back to next available engine
• If all fail, respond with text (breaking protocol, but informative)

Transcription Failure:

• Use audio metadata (duration, format)
• Generic response: "Received your audio message"

Label Generation Failure:

• Default label: [Response] [{duration}s] Audio reply

Performance Notes

Speed:

• espeak: ~1x realtime (fastest)
• OpenAI TTS: ~2-3s API call
• ElevenLabs: ~3-5s API call
• Google TTS: ~2-4s with internet

Quality:

• ElevenLabs: Best (human-like)
• OpenAI TTS: Very good
• Google TTS: Good
• espeak: Basic (robotic)

Configuration

Environment Variables:

export ELEVENLABS_API_KEY="..."
export OPENAI_API_KEY="..."
export TTS_ENGINE="elevenlabs"  # or openai, google, espeak

Config File (~/.config/audio-response.json):

{
  "tts_engine": "espeak",
  "label_format": "[{topic}] [{duration}s] {key_point}",
  "fallback_engines": ["openai", "google", "espeak"],
  "max_duration": 30
}

Testing the Protocol

Send an audio message to test:

1. System transcribes it
2. Processes request
3. Generates audio response
4. Sends with descriptive label

The first audio response will explain the TTS setup status.