Add PicoAudio Model #249

models/temporally_controllable_tta/README.md

@@ -0,0 +1,63 @@
+# PicoAudio: Enabling Precise Timing and Frequency Controllability of Audio Events in Text-to-audio Generation
+Duplicate of [github repo](https://github.com/zeyuxie29/PicoAudio)
+[![arXiv](https://img.shields.io/badge/arXiv-2407.02869v2-brightgreen.svg?style=flat-square)](https://arxiv.org/abs/2407.02869v2)
+[![githubio](https://img.shields.io/badge/GitHub.io-Audio_Samples-blue?logo=Github&style=flat-square)](https://zeyuxie29.github.io/PicoAudio.github.io/)
+[![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/ZeyuXie/PicoAudio)
+[![Hugging Face data](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Dataset-blue)](https://huggingface.co/datasets/ZeyuXie/PicoAudio/tree/main)
+
+**Bullet contribution**:
+* A data simulation pipeline tailored specifically for controllable audio generation frameworks;
+* Propose a timing-controllable audio generation framework, enabling precise control over the timing and frequency of sound event;
+* Achieve any precise control related to timing by integrating of large language models.
+
+## Inference
+You can see the demo on the website [Huggingface Online Inference](https://huggingface.co/spaces/ZeyuXie/PicoAudio) and [Github Demo](https://zeyuxie29.github.io/PicoAudio.github.io).
+Or you can use the *"inference.py"* script provided by website [Huggingface Inference](https://huggingface.co/spaces/ZeyuXie/PicoAudio/tree/main) to generate.
+Huggingface Online Inference uses Gemini as a preprocessor, and we also provide a GPT preprocessing script consistent with the paper in *"llm_preprocess.py"*
+
+## Simulated Dataset
+Simulated data can be downloaded from (1) [HuggingfaceDataset](https://huggingface.co/datasets/ZeyuXie/PicoAudio/tree/main) or (2) [BaiduNetDisk](https://pan.baidu.com/s/1rGrcjtQCEYFpr3o6y9wI8Q?pwd=pico) with the extraction code "pico". 
+The metadata is stored in *"data/meta_data/{}.json"*, one instance is as follows:
+```python
+{
+  "filepath": "data/multi_event_test/syn_1.wav",
+  "onoffCaption": "cat meowing at 0.5-2.0, 3.0-4.5 and whistling at 5.0-6.5 and explosion at 7.0-8.0, 8.5-9.5",
+  "frequencyCaption": "cat meowing two times and whistling one times and explosion two times"
+}
+```
+where:
+* *"filepath"* indicates the path to the audio file.  
+* *"frequencyCaption"* contains information about the occurrence frequency.
+* *"onoffCaption"* contains on- & off-set information.
+* For test file *"test-frequency-control_onoffFromGpt_{}.json"*, the *"onoffCaption"* is derived from *"frequencyCaption"* transformed by GPT-4, which is used for evaluation in the frequency control task.
+
+## Training 
+Download data into the *"data"* folder. 
+The training and inference code can be found in the *"picoaudio"* folder.
+```shell
+cd picoaudio
+pip install -r requirements.txt
+```
+To start traning:
+```python
+  accelerate launch runner/controllable_train.py
+```
+
+## Acknowledgement
+Our code referred to the [AudioLDM](https://github.com/haoheliu/AudioLDM) and [Tango](https://github.com/declare-lab/tango). We appreciate their open-sourcing of their code.
+
+<!--
+### Hi there 👋
+**PicoAudio/PicoAudio** is a ✨ _special_ ✨ repository because its `README.md` (this file) appears on your GitHub profile.
+
+Here are some ideas to get you started:
+
+- 🔭 I’m currently working on ...
+- 🌱 I’m currently learning ...
+- 👯 I’m looking to collaborate on ...
+- 🤔 I’m looking for help with ...
+- 💬 Ask me about ...
+- 📫 How to reach me: ...
+- 😄 Pronouns: ...
+- ⚡ Fun fact: ...
+-->

models/temporally_controllable_tta/picoaudio/audioldm/__init__.py

@@ -0,0 +1,8 @@
+from .ldm import LatentDiffusion
+from .utils import seed_everything, save_wave, get_time, get_duration
+from .pipeline import *
+
+
+
+
+

models/temporally_controllable_tta/picoaudio/audioldm/__main__.py

@@ -0,0 +1,183 @@
+#!/usr/bin/python3
+import os
+from audioldm import text_to_audio, style_transfer, build_model, save_wave, get_time, round_up_duration, get_duration
+import argparse
+
+CACHE_DIR = os.getenv(
+    "AUDIOLDM_CACHE_DIR",
+    os.path.join(os.path.expanduser("~"), ".cache/audioldm"))
+
+parser = argparse.ArgumentParser()
+
+parser.add_argument(
+    "--mode",
+    type=str,
+    required=False,
+    default="generation",
+    help="generation: text-to-audio generation; transfer: style transfer",
+    choices=["generation", "transfer"]
+)
+
+parser.add_argument(
+    "-t",
+    "--text",
+    type=str,
+    required=False,
+    default="",
+    help="Text prompt to the model for audio generation",
+)
+
+parser.add_argument(
+    "-f",
+    "--file_path",
+    type=str,
+    required=False,
+    default=None,
+    help="(--mode transfer): Original audio file for style transfer; Or (--mode generation): the guidance audio file for generating simialr audio",
+)
+
+parser.add_argument(
+    "--transfer_strength",
+    type=float,
+    required=False,
+    default=0.5,
+    help="A value between 0 and 1. 0 means original audio without transfer, 1 means completely transfer to the audio indicated by text",
+)
+
+parser.add_argument(
+    "-s",
+    "--save_path",
+    type=str,
+    required=False,
+    help="The path to save model output",
+    default="./output",
+)
+
+parser.add_argument(
+    "--model_name",
+    type=str,
+    required=False,
+    help="The checkpoint you gonna use",
+    default="audioldm-s-full",
+    choices=["audioldm-s-full", "audioldm-l-full", "audioldm-s-full-v2"]
+)
+
+parser.add_argument(
+    "-ckpt",
+    "--ckpt_path",
+    type=str,
+    required=False,
+    help="The path to the pretrained .ckpt model",
+    default=None,
+)
+
+parser.add_argument(
+    "-b",
+    "--batchsize",
+    type=int,
+    required=False,
+    default=1,
+    help="Generate how many samples at the same time",
+)
+
+parser.add_argument(
+    "--ddim_steps",
+    type=int,
+    required=False,
+    default=200,
+    help="The sampling step for DDIM",
+)
+
+parser.add_argument(
+    "-gs",
+    "--guidance_scale",
+    type=float,
+    required=False,
+    default=2.5,
+    help="Guidance scale (Large => better quality and relavancy to text; Small => better diversity)",
+)
+
+parser.add_argument(
+    "-dur",
+    "--duration",
+    type=float,
+    required=False,
+    default=10.0,
+    help="The duration of the samples",
+)
+
+parser.add_argument(
+    "-n",
+    "--n_candidate_gen_per_text",
+    type=int,
+    required=False,
+    default=3,
+    help="Automatic quality control. This number control the number of candidates (e.g., generate three audios and choose the best to show you). A Larger value usually lead to better quality with heavier computation",
+)
+
+parser.add_argument(
+    "--seed",
+    type=int,
+    required=False,
+    default=42,
+    help="Change this value (any integer number) will lead to a different generation result.",
+)
+
+args = parser.parse_args()
+
+if(args.ckpt_path is not None):
+    print("Warning: ckpt_path has no effect after version 0.0.20.")
+
+assert args.duration % 2.5 == 0, "Duration must be a multiple of 2.5"
+
+mode = args.mode
+if(mode == "generation" and args.file_path is not None):
+    mode = "generation_audio_to_audio"
+    if(len(args.text) > 0):
+        print("Warning: You have specified the --file_path. --text will be ignored")
+        args.text = ""
+
+save_path = os.path.join(args.save_path, mode)
+
+if(args.file_path is not None):
+    save_path = os.path.join(save_path, os.path.basename(args.file_path.split(".")[0]))
+
+text = args.text
+random_seed = args.seed
+duration = args.duration
+guidance_scale = args.guidance_scale
+n_candidate_gen_per_text = args.n_candidate_gen_per_text
+
+os.makedirs(save_path, exist_ok=True)
+audioldm = build_model(model_name=args.model_name)
+
+if(args.mode == "generation"):
+    waveform = text_to_audio(
+        audioldm,
+        text,
+        args.file_path,
+        random_seed,
+        duration=duration,
+        guidance_scale=guidance_scale,
+        ddim_steps=args.ddim_steps,
+        n_candidate_gen_per_text=n_candidate_gen_per_text,
+        batchsize=args.batchsize,
+    )
+
+elif(args.mode == "transfer"):
+    assert args.file_path is not None
+    assert os.path.exists(args.file_path), "The original audio file \'%s\' for style transfer does not exist." % args.file_path
+    waveform = style_transfer(
+        audioldm,
+        text,
+        args.file_path,
+        args.transfer_strength,
+        random_seed,
+        duration=duration,
+        guidance_scale=guidance_scale,
+        ddim_steps=args.ddim_steps,
+        batchsize=args.batchsize,
+    )
+    waveform = waveform[:,None,:]
+
+save_wave(waveform, save_path, name="%s_%s" % (get_time(), text))

models/temporally_controllable_tta/picoaudio/audioldm/audio/__init__.py

@@ -0,0 +1,2 @@
+from .tools import wav_to_fbank, read_wav_file
+from .stft import TacotronSTFT

models/temporally_controllable_tta/picoaudio/audioldm/audio/audio_processing.py

@@ -0,0 +1,100 @@
+import torch
+import numpy as np
+import librosa.util as librosa_util
+from scipy.signal import get_window
+
+
+def window_sumsquare(
+    window,
+    n_frames,
+    hop_length,
+    win_length,
+    n_fft,
+    dtype=np.float32,
+    norm=None,
+):
+    """
+    # from librosa 0.6
+    Compute the sum-square envelope of a window function at a given hop length.
+
+    This is used to estimate modulation effects induced by windowing
+    observations in short-time fourier transforms.
+
+    Parameters
+    ----------
+    window : string, tuple, number, callable, or list-like
+        Window specification, as in `get_window`
+
+    n_frames : int > 0
+        The number of analysis frames
+
+    hop_length : int > 0
+        The number of samples to advance between frames
+
+    win_length : [optional]
+        The length of the window function.  By default, this matches `n_fft`.
+
+    n_fft : int > 0
+        The length of each analysis frame.
+
+    dtype : np.dtype
+        The data type of the output
+
+    Returns
+    -------
+    wss : np.ndarray, shape=`(n_fft + hop_length * (n_frames - 1))`
+        The sum-squared envelope of the window function
+    """
+    if win_length is None:
+        win_length = n_fft
+
+    n = n_fft + hop_length * (n_frames - 1)
+    x = np.zeros(n, dtype=dtype)
+
+    # Compute the squared window at the desired length
+    win_sq = get_window(window, win_length, fftbins=True)
+    win_sq = librosa_util.normalize(win_sq, norm=norm) ** 2
+    win_sq = librosa_util.pad_center(win_sq, n_fft)
+
+    # Fill the envelope
+    for i in range(n_frames):
+        sample = i * hop_length
+        x[sample : min(n, sample + n_fft)] += win_sq[: max(0, min(n_fft, n - sample))]
+    return x
+
+
+def griffin_lim(magnitudes, stft_fn, n_iters=30):
+    """
+    PARAMS
+    ------
+    magnitudes: spectrogram magnitudes
+    stft_fn: STFT class with transform (STFT) and inverse (ISTFT) methods
+    """
+
+    angles = np.angle(np.exp(2j * np.pi * np.random.rand(*magnitudes.size())))
+    angles = angles.astype(np.float32)
+    angles = torch.autograd.Variable(torch.from_numpy(angles))
+    signal = stft_fn.inverse(magnitudes, angles).squeeze(1)
+
+    for i in range(n_iters):
+        _, angles = stft_fn.transform(signal)
+        signal = stft_fn.inverse(magnitudes, angles).squeeze(1)
+    return signal
+
+
+def dynamic_range_compression(x, normalize_fun=torch.log, C=1, clip_val=1e-5):
+    """
+    PARAMS
+    ------
+    C: compression factor
+    """
+    return normalize_fun(torch.clamp(x, min=clip_val) * C)
+
+
+def dynamic_range_decompression(x, C=1):
+    """
+    PARAMS
+    ------
+    C: compression factor used to compress
+    """
+    return torch.exp(x) / C