Text-to-Speech (TTS)

The Text-to-Speech Layer is the last stage in the /audio-chat pipeline. After generating and optionally refining the answer, Billx-Agent uses ElevenLabs to convert that text into a natural-sounding voice response.

---

🎯 What It Does

• Accepts a text string (e.g. "Top 5 products are...")

• Sends it to ElevenLabs TTS API

• Receives an audio file in return

• Encodes the audio in Base64 format

• Includes the audio in the API response so the client can play it

---

🔉 Where It's Used

• Primarily in POST /audio-chat

• Can also be used standalone via POST /tts

---

🔁 Audio Response Example

{
  "refined_answer": "The top 3 selling products are A, B, and C.",
  "audio_content": "<base64-encoded-audio>"
}

You can play the audio in your frontend using JavaScript like this:

const audio = new Audio("data:audio/mp3;base64,<audio_content>");
audio.play();

---

⚙️ Output Format

• Default format: MP3 (via ElevenLabs)

• Returned as base64 string in API response

• Compatible with web, mobile, and desktop audio players

---

🎙️ Voice Customization (Advanced)

If supported by your ElevenLabs plan, you can:

• Choose different voices

• Adjust speech rate and pitch

• Localize for different languages (future-ready)

---

📌 TTS is optional — if your users prefer reading results, you can ignore the audio_content field.

---