murmr 0.4.0

Python SDK for the murmr TTS API

murmr

Python SDK for the murmr TTS API. Async-first with full sync support.

pip install murmr

Quick Start

Voice Design (describe any voice in natural language)

from murmr import MurmrClient

client = MurmrClient(api_key="murmr_sk_live_...")

# Generate speech with a voice description
wav = client.voices.design(
    input="Hello, welcome to murmr!",
    voice_description="A warm, friendly female voice with a slight British accent",
)

with open("output.wav", "wb") as f:
    f.write(wav)

Saved Voices (batch via RunPod Serverless)

# Submit a batch job
job = client.speech.create(input="Hello world", voice="voice_abc123")

# Wait for completion
result = client.speech.create_and_wait(input="Hello world", voice="voice_abc123")
audio = result.audio_bytes  # decoded WAV

Streaming

# Stream PCM audio chunks
with client.speech.stream(input="Hello world", voice="voice_abc123") as stream:
    for chunk in stream:
        pcm = chunk.audio_bytes  # 24kHz mono 16-bit PCM
        if chunk.done:
            break

Async

import asyncio
from murmr import AsyncMurmrClient

async def main():
    async with AsyncMurmrClient(api_key="murmr_sk_live_...") as client:
        wav = await client.voices.design(
            input="Hello from async!",
            voice_description="A deep male voice",
        )

asyncio.run(main())

Long-Form Audio

# Automatically chunks, retries, and concatenates
result = client.speech.create_long_form(
    input=very_long_text,
    voice="voice_abc123",
    on_progress=lambda current, total, pct: print(f"{pct}%"),
)

with open("long_form.wav", "wb") as f:
    f.write(result.audio)

API Reference

Clients

Class	Description
MurmrClient(api_key=...)	Sync client (context manager)
AsyncMurmrClient(api_key=...)	Async client (async context manager)

Speech (client.speech)

Method	Returns	Description
create(input, voice, ...)	AsyncJobResponse	Submit batch job
create_and_wait(input, voice, ...)	JobStatus	Submit and poll until done
stream(input, voice, ...)	Context manager yielding AudioStreamChunk	Stream PCM chunks
create_long_form(input, voice, ...)	LongFormResult	Chunk + concat long text

Voices (client.voices)

Method	Returns	Description
design(input, voice_description, ...)	bytes (WAV)	Generate with voice description
design_stream(input, voice_description, ...)	Context manager yielding AudioStreamChunk	Stream voice design

Jobs (client.jobs)

Method	Returns	Description
get(job_id)	JobStatus	Get job status
wait_for_completion(job_id, ...)	JobStatus	Poll until done/failed

Text Formatting

Newline characters in your input text affect prosody:

• \n (single newline) creates a sentence-level breath pause
• \n\n (double newline) creates a paragraph-level pause with prosodic reset
• No newlines in long text produces rushed, flat delivery

Best practice: Insert \n between sentences and \n\n between paragraphs. Avoid text with hard line wraps every 60-80 characters (e.g., from PDFs or terminals) — this produces choppy output.

See the Text Formatting Guide for details and preprocessing examples.

Supported Languages

Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian

License

MIT