Upload 4 files

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 anttiluode
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,15 +1,25 @@
----
-title: BreakSongToPCAModes
-emoji: 🌖
-colorFrom: gray
-colorTo: blue
-sdk: gradio
-sdk_version: 6.26.0
-python_version: '3.12'
-app_file: app.py
-pinned: false
-license: mit
-short_description: 'Breaks songs to PCA modes that may be accepted by Suno. '
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+---
+title: Break Song To PCA Modes
+emoji: 🎵
+colorFrom: indigo
+colorTo: purple
+sdk: gradio
+sdk_version: 4.44.0
+app_file: app.py
+pinned: false
+license: mit
+short_description: Decompose audio tracks into PCA modes.
+---
+
+# Break Song To PCA Modes
+
+An interactive audio decomposition app that breaks down songs into PCA modes.
+
+## Installation & Local Run
+
+```bash
+git clone [https://huggingface.co/spaces/](https://huggingface.co/spaces/)<your-username>/BreakSongToPCAModes
+cd BreakSongToPCAModes
+pip install -r requirements.txt
+python app.py
+```

app.py

@@ -0,0 +1 @@
+#!/usr/bin/env python3
    L = (dim - 1) * tau + 1
    win = sliding_window_view(x, L)[:, ::tau]
    return win[:, ::-1].copy()
    """
    Returns list of raw PCA mode signals (time domain) of length same as input.
    Each mode is a reconstructed time series from one singular vector.
    """
    # Delay-embed
    X = embed(x, dim, tau)
    # Normalise rows
    a = np.linalg.norm(X, axis=1) + 1e-9
    Xv = X / a[:, None]
    
    # Randomized SVD
    U, S, Vt = randomized_svd(Xv, n_components=k, n_iter=5, random_state=0)
    W = Vt  # shape (k, dim)
    
    # Project each mode back to original space
    Y = Xv @ W.T  # activations (T x k)
    
    raw_modes = []
    for i in range(k):
        # Reconstruct this mode alone
        mode_embedded = Y[:, i:i+1] @ W[i:i+1, :]  # (T, dim)
        mode_signal = np.real(mode_embedded[:, 0]) * a
        raw_modes.append(mode_signal)
    
    return raw_modes
    if file is None:
        raise gr.Error("Please upload an audio file.")
    
    # Load and preprocess
    try:
        sig, fs = sf.read(file)
    except Exception as e:
        raise gr.Error(f"Failed to read: {e}")
    
    if sig.ndim > 1:
        sig = np.mean(sig, axis=1)
    
    # Resample to 16 kHz for speed/consistency
    target_sr = 16000
    if fs != target_sr:
        sig = signal.resample_poly(sig, target_sr, fs)
        fs = target_sr
    
    # Normalise
    sig = sig / (np.max(np.abs(sig)) + 1e-9)
    
    duration = len(sig) / fs
    gr.Info(f"Processing {duration:.1f} sec song with dim={dim}, tau={tau}, k={k} ...")
    
    # Extract raw modes
    try:
        raw_modes = extract_raw_modes(sig, dim=dim, tau=tau, k=k)
    except Exception as e:
        raise gr.Error(f"Extraction failed: {e}")
    
    # Create ZIP with raw mode WAVs
    temp_dir = tempfile.mkdtemp()
    for i, mode in enumerate(raw_modes):
        sf.write(os.path.join(temp_dir, f"raw_mode_{i:02d}.wav"), mode, fs)
    
    zip_path = os.path.join(tempfile.gettempdir(), "raw_modes.zip")
    with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zf:
        for fname in os.listdir(temp_dir):
            zf.write(os.path.join(temp_dir, fname), arcname=fname)
    
    # Also preview the first raw mode (optional)
    preview_audio = (fs, raw_modes[0])
    return preview_audio, zip_path
    gr.Markdown("""
    # 🔧 Fast Raw Mode Extractor (GN‑v7, PCA only)
    Upload your song → get raw Koopman/PCA modes as WAV files.
    **No ICA, no plots – just the raw modes.**  
    These are the `raw_mode_XX.wav` files from the original GN‑v7.
    """)
    with gr.Row():
        with gr.Column():
            audio_in = gr.Audio(type="filepath", label="Your song (any length)")
            dim_slider = gr.Slider(16, 64, value=32, step=2, label="Embedding dim (lower = faster)")
            tau_slider = gr.Slider(1, 8, value=3, step=1, label="Delay tau")
            k_slider = gr.Slider(2, 16, value=8, step=1, label="Number of raw modes (k)")
            run_btn = gr.Button("Extract Raw Modes", variant="primary")
        with gr.Column():
            preview = gr.Audio(label="Preview of first raw mode", type="numpy")
            zip_output = gr.File(label="📦 Download all raw modes (ZIP)")
    
    run_btn.click(
        fn=process_to_raw_modes,
        inputs=[audio_in, dim_slider, tau_slider, k_slider],
        outputs=[preview, zip_output]
    )
    demo.launch()

requirements.txt

@@ -0,0 +1,5 @@
+gradio>=3.50.0
+numpy>=1.24.0
+scipy>=1.10.0
+soundfile>=0.12.0
+scikit-learn>=1.3.0