Voice AI · Python

Voice Conversation Pipeline

Build end-to-end voice AI: speech recognition, LLM reasoning, and text-to-speech in a single pipeline. Sub-200ms latency.

1. Install

pip install xalen

2. Code

# 1. Transcribe user speech
transcript = client.audio.transcriptions.create(
    model=#A6E3A1;">"whisper-large-v3",
    file=open(#A6E3A1;">"user_question.mp3", "rb"),
    language=#A6E3A1;">"hi"
)

# 2. AI reasoning
response = client.chat.completions.create(
    model=#A6E3A1;">"vedika-fast",  # Low latency
    messages=[{#A6E3A1;">"role": "user", "content": transcript.text}]
)

# 3. Speak the response
audio = client.audio.speech.create(
    model=#A6E3A1;">"vedika-pandit-voice",
    input=response.choices[0].message.content,
    language=#A6E3A1;">"hi"
)
audio.stream_to_file(#A6E3A1;">"response.mp3")

Related Tutorials

Last updated: 2026-05-21