lexigram-multimedia-beat

Audio tempo/beat analysis for the Lexigram Framework — analyzes an audio MediaAsset and returns tempo (BPM) and beat timestamps for driving beat-synced cut timing.

Overview

lexigram-multimedia-beat analyzes an audio clip and returns a BeatAnalysisResult (tempo_bpm + beat_timestamps). Two backends are available: librosa (default, runs in-process, no reference server) and madmom (optional, deep-learning-based, reference-server pattern — better on syncopated or tempo-changing material).

Full documentation: docs.lexigram.dev

Install

uv add lexigram-multimedia-beat
# Optional extras
uv add "lexigram-multimedia-beat[librosa]"        # librosa in-process backend
uv add "lexigram-multimedia-beat[madmom-server]"  # madmom reference server

Quick Start

from lexigram import Application
from lexigram.di.module import Module, module
from lexigram.multimedia.beat import BeatAnalysisModule
from lexigram.contracts.multimedia.protocols import BeatAnalysisProvider
from lexigram.contracts.multimedia.types import BeatAnalysisRequest, MediaAsset


@module(imports=[BeatAnalysisModule.configure()])
class AppModule(Module):
    pass


async def main() -> None:
    async with Application.boot(modules=[AppModule]) as app:
        beat = await app.container.resolve(BeatAnalysisProvider)
        asset = MediaAsset(mime_type="audio/mp3", provider="local", bytes_data=b"<mp3>")
        result = await beat.analyze(BeatAnalysisRequest(asset=asset))
        if result.is_ok():
            analysis = result.unwrap()  # BeatAnalysisResult — tempo_bpm + beat_timestamps
            print(analysis.tempo_bpm)


if __name__ == "__main__":
    import asyncio

    asyncio.run(main())

Configuration

Zero-config usage: Call BeatAnalysisModule.configure() with no arguments to use the in-process librosa backend.

Option 1 — YAML file

# application.yaml
multimedia_beat:
  backend: "librosa"
  librosa_sample_rate: 44100

Option 2 — Profiles + Environment Variables

export LEX_PROFILE=production
export LEX_MULTIMEDIA_BEAT__BACKEND=madmom
export LEX_MULTIMEDIA_BEAT__LIBROSA_SAMPLE_RATE=44100

Option 3 — Python

from lexigram.multimedia.beat import BeatAnalysisModule
from lexigram.multimedia.beat.config import BeatAnalysisConfig

BeatAnalysisModule.configure(
    config=BeatAnalysisConfig(backend="librosa", librosa_sample_rate=44100)
)

Config reference

Module Factory Methods

Key Features

• Two backends — librosa (in-process, default, no reference server) and madmom (deep-learning, reference-server pattern)
• No blob persistence — returns a BeatAnalysisResult (tempo + beat timestamps), not a MediaAsset; no queued task handler
• Beat-synced editing — drive cut timing in the calling application
• Reference server — lexigram-beat-madmom-serve console script runs the madmom model server
• Result-based — analyze() -> Result[BeatAnalysisResult, MultimediaError]; errors are domain values, not exceptions

Testing

from lexigram import Application
from lexigram.multimedia.beat import BeatAnalysisModule

async def test_boot():
    async with Application.boot(modules=[BeatAnalysisModule.stub()]) as app:
        assert app.container is not None

Key Source Files