OpenFLAM

Joint Audio and Text Embeddings via Framewise Language-Audio Modeling (FLAM)

FLAM is a cutting-edge language–audio model that supports both zero-shot sound even detection and large-scale audio retrieval via free-form text.

This code accompanies the following ICML 2025 publication:

@inproceedings{flam2025,
  title = {FLAM: Frame-Wise Language-Audio Modeling},
  author = {Yusong Wu and Christos Tsirigotis and Ke Chen and Cheng-Zhi Anna Huang and Aaron Courville and Oriol Nieto and Prem Seetharaman and Justin Salamon},
  booktitle={International Conference on Machine Learning, ICML},
  year = {2025}
}

Architecture

FLAM is based on contrastive language-audio pretraining, known as CLAP, and improve its capability by supporting the frame-wise event localization via learnable text and audio biases and scales.

Quick Start

Install FLAM via PyPi:

pip install openflam

Two examples are provided:

1. embedding_inference.py: to obtain audio and text embeddings and do sound event localization.
2. sed_inference_and_plot.py to do sound event localization and plot the results.

For the API documentation, please refer to hook.py.

To obtain audio and text embeddings

Please refer to embedding_inference.py:

import os
import librosa
import openflam
import torch


DEVICE = "cuda"  # cuda or cpu
SR = 48000       # Sampling Rate (FLAM requires 48kHz)

flam = openflam.OpenFLAM(
    model_name="v1-base", default_ckpt_path="/tmp/openflam"
).to(DEVICE)

# Sanity Check (Optional)
flam.sanity_check()

# load audio from 22-33 seconds
audio, sr = librosa.load("test/test_data/test_example.mp3", sr=SR)
audio = audio[int(23. * sr): int(33. * sr)]
audio_samples = torch.tensor(audio).unsqueeze(0).to(DEVICE) # [B, 480000 = 10 sec]

# Define text
text_samples = [
    "man speaking",
    "man talking through a walkie-talkie",
    "music",
    "breathing sound",
    "ratcheting"
]

# Get Global Audio Features (10sec = 0.1Hz embeddings)
audio_global_feature = flam.get_global_audio_features(
    audio_samples
)  # [B, 512]

# Get Local Audio Features (0.32sec = ~3Hz embeddings)
audio_local_feature = flam.get_local_audio_features(
    audio_samples
)  # [B, 32, 512] 32 is frame size (0.032 sec / frame)

# Get Text Features
text_feature = flam.get_text_features(text_samples)  # [B, 512]

# Get Local Similarity for Sound Event Detection
flamgram = flam.get_local_similarity(
    audio_samples,
    text_samples,
    method="unbiased",
    cross_product=True,
)

To perform sound event localization and plot the diagram

Please refer to sed_inference_and_plot.py.

You should be able to see such plot by running the below codes:

import torch
import numpy as np
import librosa
import scipy
from pathlib import Path
import openflam
from openflam.module.plot_utils import plot_sed_heatmap

TEXTS = [
    "man speaking",
    "man talking through a walkie-talkie",
    "music",
    "breathing sound",
    "ratcheting",
]

NEGATIVE_CLASS = [
    "ratcheting",
]

flam_wrapper = openflam.OpenFLAM(
      model_name="v1-base", default_ckpt_path="/tmp/openflam"
)
flam_wrapper.to("cuda")

# Load and prepare audio
audio, sr = librosa.load("test_data/test_example.mp3", sr=48000)
audio = audio[int(22. * sr) : int(33. * sr)]

# Convert to tensor and move to device
audio_tensor = torch.tensor(audio).unsqueeze(0).to("cuda")

# Run inference
with torch.no_grad():
    # Get local similarity using the wrapper's built-in method
    # This uses the unbiased method (Eq. 9 in the paper)
    act_map_cross = (
        flam_wrapper.get_local_similarity(
            audio_tensor,
            TEXTS,
            method="unbiased",
            cross_product=True,
        )
        .cpu()
        .numpy()
    )

# Apply median filtering for smoother results
act_map_filter = []
for i in range(act_map_cross.shape[0]):
    act_map_filter.append(
        scipy.ndimage.median_filter(act_map_cross[i], (1, 3))
    )
act_map_filter = np.array(act_map_filter)

# Prepare similarity dictionary for plotting
similarity = {
    f"{TEXTS[i]}": act_map_filter[0][i] for i in range(len(TEXTS))
}

# Prepare audio for plotting (resample to 32kHz)
audio_plot = librosa.resample(
    audio, orig_sr=48000, target_sr=32000
)

# Generate and save visualization
output_path = "sed_output/sed_heatmap_22s-33s.png"
plot_sed_heatmap(
    audio_plot,
    32000,
    post_similarity=similarity,
    duration=10.0,
    negative_class=NEGATIVE_CLASS,
    figsize=(14, 8),
    save_path=output_path,
)

License

Both code and models for OpenFLAM are released under a non-commercial Adobe Research License. Please, review it carefully before using this technology.

Pretrained Models

The pretrained checkpoints can be found here.

OpenFLAM automatically handles the downloading of the checkpoint. Please, refer to the previous section for more details.

Datasets

The original experimental results reported in our paper were obtained by the model trained on internal datasets that are not publicly shareable.

OpenFLAM is trained on all publicly available datasets, including:

1. Datasets with coarse (aka, global or weak) labels: AudioSet-ACD (a LLM-based captioning for AudioSet), FreeSound, WavCaps, AudioCaps, Clotho;
2. Datasets with fine-grained (aka, local or strong) labels: AudioSet Strong, UrbanSED, DESED, Maestro, and Simulation data from AudioSet-ACD & FreeSound.

We report a comparison of the OpenFLAM performance to the original paper report (the global retrieval metrics --ie, A2T and T2A-- are R@1 / R@5):

Citation

If you use OpenFLAM, please cite our main work:

@inproceedings{flam2025,
  title = {FLAM: Frame-Wise Language-Audio Modeling},
  author = {Yusong Wu and Christos Tsirigotis and Ke Chen and Cheng-Zhi Anna Huang and Aaron Courville and Oriol Nieto and Prem Seetharaman and Justin Salamon},
  booktitle={International Conference on Machine Learning, ICML},
  year = {2025}
}