Arctan

Real-time speech enhancement and noise cancellation using Arctan Eigen

Overview

ArctanAudioFilter is a BaseAudioFilter implementation backed by Arctan Eigen, a low-latency speech enhancer and noise canceller tuned for ASR/STT accuracy. You attach it to a transport’s audio_in_filter, and it cleans up incoming user audio — reducing background noise, side-talk, and other degradation — before the audio reaches VAD and your STT service.

Arctan Eigen Docs

SDK documentation, API reference, and troubleshooting

PyPI Package

The arctan-vi package on PyPI

Pipecat Integration Guide

Arctan’s own guide to the Pipecat integration

Request Access

Request an SDK license key from Arctan

Installation

This is a community-maintained package distributed separately from pipecat-ai:

uv add "arctan-vi[pipecat]"

The pipecat extra installs the Arctan SDK along with the Pipecat integration.

Requires Python >= 3.11.

Prerequisites

• SDK license key: Arctan requires a license key. Request access at arctan.ai/eigen.
• Network access: the agent process must be able to reach Arctan’s servers to validate the license key.

Required Environment Variables

• ARCTAN_SDK_KEY: your Arctan SDK license key

export ARCTAN_SDK_KEY="arc_sk_live_..."

Don’t commit SDK license keys or .env files to source control.

Configuration

license_key · str | None · default: "None"

Arctan SDK license key. Falls back to the ARCTAN_SDK_KEY environment variable if not provided. When set, this value takes precedence over the environment variable.

Usage

Create one filter instance and pass it to your transport as audio_in_filter:

from arctan.pipecat import arctan_enhancer
from pipecat.transports.base_transport import TransportParams

audio_filter = arctan_enhancer.ArctanAudioFilter()

transport_params = TransportParams(
    audio_in_enabled=True,
    audio_in_filter=audio_filter,
)

The transport drives the filter lifecycle — start(sample_rate), filter(audio), process_frame(frame), and stop() — so no additional wiring is needed.

Passing the license key explicitly

audio_filter = arctan_enhancer.ArctanAudioFilter(
    license_key="arc_sk_live_...",
)

With a specific transport

The filter works with any Pipecat transport that supports input audio filters:

from arctan.pipecat import arctan_enhancer
from pipecat.transports.daily.transport import DailyParams, DailyTransport

audio_filter = arctan_enhancer.ArctanAudioFilter()

transport = DailyTransport(
    room_url,
    token,
    "Bot",
    DailyParams(
        audio_in_enabled=True,
        audio_out_enabled=True,
        audio_in_filter=audio_filter,
    ),
)

Input Frames

FilterEnableFrame · Frame

Control frame to toggle enhancement on and off at runtime

from pipecat.frames.frames import FilterEnableFrame

# Bypass enhancement — audio passes through unchanged
await task.queue_frame(FilterEnableFrame(enable=False))

# Resume enhancement
await task.queue_frame(FilterEnableFrame(enable=True))

Audio Requirements

The Arctan filter does not resample audio. Configure your input transport to deliver PCM16 at a single, stable sample rate.

• Input must be signed 16-bit PCM (int16) bytes.
• The sample rate must be positive and must not change after the filter starts.
• Each chunk passed to filter() must contain complete int16 samples; incomplete samples raise an error.
• Any chunk size is accepted — the filter buffers partial data internally.

Compatibility

Requires Pipecat 1.3 or later. Check the Arctan documentation for the latest tested version and release notes.