Livekit Murf
Murf AI text-to-speech and speech-to-text for Livekit Agents.
Note: This integration is maintained by Murf AI.
Use this package for both speech and transcription. murf.TTS follows the same Murf requests as LiveKit's Murf TTS plugin: Falcon by default, api-key header, and the same voice, style, speed, pitch, and locale fields. Do not install livekit-plugins-murf in the same environment. Both packages provide livekit.plugins.murf, and the later install replaces the other.
Installation
pip install livekit-murf
TTS and STT only (pure Python). For background voice cancellation on the agent's audio input:
pip install "livekit-murf[bvc]"
That pulls the platform-specific livekit-plugins-murf-bvc wheel (native libmurf_bvc for your OS/arch). Pip downloads only that platform's wheel.
from livekit.plugins import murf
session = AgentSession(
stt=murf.STT(model="finch_indic_1.0", language="hi,te"),
tts=murf.TTS(voice="Pooja", style="Conversational", model="FALCON"),
)
Background voice cancellation (optional)
Requires pip install "livekit-murf[bvc]", a Murf API key with BVC enabled (MURF_API_KEY), and Linux (glibc 2.28+) or macOS.
from livekit.agents import Agent, AgentServer, AgentSession, JobContext, JobProcess, cli, room_io
from livekit.plugins import murf, murf_bvc
server = AgentServer()
def prewarm(proc: JobProcess):
murf_bvc.prewarm()
server.setup_fnc = prewarm
@server.rtc_session()
async def entrypoint(ctx: JobContext):
session = AgentSession(
stt=murf.STT(model="finch_global_1.0"),
tts=murf.TTS(voice="Natalie", style="Conversational", model="FALCON"),
)
await session.start(
room=ctx.room,
agent=Agent(instructions="You are a helpful assistant."),
room_options=room_io.RoomOptions(
audio_input=room_io.AudioInputOptions(noise_cancellation=murf_bvc.BVC()),
),
)
if __name__ == "__main__":
cli.run_app(server)
Livekit Starter Project
from livekit.agents import JobContext, WorkerOptions, cli, JobProcess, get_job_context
from livekit.agents.voice import Agent, AgentSession
from livekit.plugins import murf, openai, silero
class MyAgent(Agent):
def __init__(self) -> None:
super().__init__(
instructions="You are a voice agent build using Murf",
stt=murf.STT(model="finch_indic_1.0", language="hi,te"),
llm=openai.LLM(model="gpt-4o"),
tts=murf.TTS(voice="Pooja", style="Conversational", model="FALCON"),
vad=get_job_context().proc.userdata["vad"],
)
async def on_enter(self):
await self.session.say("Hi, I am a voice agent powered by Murf, how can I help you?")
async def entrypoint(ctx: JobContext):
await ctx.connect()
ctx.log_context_fields = {
"room": ctx.room.name,
}
session = AgentSession()
await session.start(
agent=MyAgent(),
room=ctx.room
)
def prewarm(proc: JobProcess):
proc.userdata["vad"] = silero.VAD.load()
if __name__ == "__main__":
cli.run_app(WorkerOptions(entrypoint_fnc=entrypoint, prewarm_fnc=prewarm))
Speech to Text
murf.STT streams PCM audio to Murf's streaming transcription WebSocket. Partials are interim transcripts. Murf sends a final when the caller pauses. LiveKit's turn detection decides when the agent replies.
stt = murf.STT(model="finch_indic_1.0", language="hi,te")
Parameters
model(optional) — Default:finch_global_1.0.finch_global_1.0(us-east.api.murf.ai) transcribes English and European languages.finch_indic_1.0(in.api.murf.ai) transcribes English (India) and Indic languages.language(optional) — Default:auto, which detects among the languages of the connected model. A comma-separated string or a list of codes, such as"hi,te"or["hi", "te"], detects only from those languages. BCP-47 tags such asen-USorhi-INare mapped to Murf's codes.sample_rate(optional) — Default:16000. Incoming audio is resampled to this rate. One of8000,12000,16000,24000,32000,44100, or48000.keywords(optional) — Words or phrases to boost. Sent as Murf's comma-separatedkeywordsquery parameter.alpha(optional) — Default:0.5. How strongly to favourkeywords, from0.0to1.0.text_formatting(optional) — Default:False. Format numbers and similar terms. Applies to Hindi and English (India) only.bvc(optional) — Default:False. Remove background voices before transcription.
The API key is read from MURF_API_KEY when api_key is omitted.
Text to Speech
murf.TTS synthesizes speech with Murf. It uses the same request fields as LiveKit's Murf TTS plugin. Audio streams over a WebSocket. Set streaming=False to synthesize over HTTP instead.
tts = murf.TTS(
voice="Pooja",
style="Conversational",
model="FALCON",
)
Parameters
model(optional) — Default:FALCON. Murf TTS model. Valid values areFALCONandGEN2.voice(optional) — Default:Gordon. Voice ID from Murf's voice library.style(optional) — Default:Conversation. Voice style, for exampleConversational,Narration, orPromo. See the voice library for the styles each voice supports.locale(optional) — Locale for synthesis, for exampleen-USorhi-IN. When omitted, the locale is inferred from the voice.speed(optional) — Speech speed. Positive values are faster and negative values are slower. Valid range:-50to50.pitch(optional) — Speech pitch. Positive values are higher and negative values are lower. Valid range:-50to50.
sample_rate defaults to 24000 and encoding defaults to pcm. The API key is read from MURF_API_KEY when api_key is omitted. update_options() changes locale, voice, style, speed, and pitch for later audio.
Pre-requisites
You'll need an API key from Murf AI. It can be set as an environment variable: MURF_API_KEY
Compatibility: Requires livekit-agents >= 1.5.2.
Metadata
Release files for livekit-murf 0.2.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| livekit_murf-0.2.1.tar.gz | 20.4 kB | Without provenance | Details for livekit_murf-0.2.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| livekit_murf-0.2.1-py3-none-any.whl | Python 3 | none | any | Details for livekit_murf-0.2.1-py3-none-any.whl |
Total release size: 41.3 kB
Release files / livekit_murf-0.2.1.tar.gz
| Download URL | livekit_murf-0.2.1.tar.gz |
|---|---|
| Size | 20.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
1b896fbd8d5195ea413534255186a010bca22168b7456523e1fd4a2138932cbd
|
|
BLAKE2b-256 checksum How to use checksums |
b4bf855c85c8667808511b2e6a125f2221f030d054112ac80b08c89117f2c5d2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / livekit_murf-0.2.1-py3-none-any.whl
| Download URL | livekit_murf-0.2.1-py3-none-any.whl |
|---|---|
| Size | 20.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
62aaac091e8f0a44d894a4e0fd2a15da3feb6ecf3598e405e90e89da8fd47172
|
|
BLAKE2b-256 checksum How to use checksums |
1c87429880b41c390e8cb78d4a83d5f6e0203fbcb98f52827c361a4458c9bcc8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for livekit-murf 0.2.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.