inital commit

README.md

@@ -1,13 +1,14 @@
 ---
-title: Harry Potter Spells
-emoji: 🐨
-colorFrom: yellow
-colorTo: gray
+title: Spell Recorder
+emoji: ✨
+colorFrom: indigo
+colorTo: purple
+colorBottom: purple
 sdk: gradio
-sdk_version: 5.49.1
 app_file: app.py
 pinned: false
-short_description: This space is for collect Harry potter spell voices
+license: mit
+short_description: Collect microphone recordings for six spells (Lumos, Nox, Alohomora, Wingardium Leviosa, Accio, Reparo)
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,176 @@
+import os
+import re
+import time
+import math
+from typing import List, Tuple, Optional
+
+import numpy as np
+import gradio as gr
+import soundfile as sf
+from scipy.signal import resample_poly
+
+# Output directory for saved recordings
+OUT_DIR = "recordings"
+os.makedirs(OUT_DIR, exist_ok=True)
+
+# Fixed target sample rate for ML training
+TARGET_SR = 16000
+
+# Spells to collect
+SPELLS = [
+    "Lumos",
+    "Nox",
+    "Alohomora",
+    "Wingardium Leviosa",
+    "Accio",
+    "Reparo",
+]
+
+
+def sanitize_username(name: Optional[str]) -> str:
+    """Sanitize username for safe filenames.
+    - only keep a-z, 0-9, dash and underscore
+    - collapse whitespace to underscore
+    - default to 'anon' if empty
+    """
+    if not name:
+        return "anon"
+    # normalize whitespace then strip
+    name = re.sub(r"\s+", "_", name.strip())
+    # keep safe chars only
+    name = re.sub(r"[^a-zA-Z0-9_-]", "", name)
+    return name.lower() or "anon"
+
+
+def to_mono(audio: np.ndarray) -> np.ndarray:
+    if audio.ndim == 2:
+        # average channels to mono
+        return audio.mean(axis=1)
+    return audio
+
+
+def resample_to_target(audio: np.ndarray, sr: int, target_sr: int = TARGET_SR) -> np.ndarray:
+    if sr == target_sr:
+        return audio
+    # rational resampling factors
+    g = math.gcd(sr, target_sr)
+    up = target_sr // g
+    down = sr // g
+    return resample_poly(audio, up=up, down=down)
+
+
+def save_one_from_path(filepath: Optional[str], spell: str, username: str) -> Optional[str]:
+    """Load an audio file path (from mic/upload), process to 16k mono, and save.
+    Returns saved file path or None if no audio provided.
+    """
+    if not filepath:
+        return None
+
+    audio, sr = sf.read(filepath, dtype="float32", always_2d=False)
+    if audio is None or (isinstance(audio, np.ndarray) and audio.size == 0):
+        return None
+
+    audio = to_mono(np.asarray(audio))
+    audio = resample_to_target(audio, sr, TARGET_SR)
+    audio = np.clip(audio, -1.0, 1.0)
+
+    # Build descriptive filename: spell_username_timestamp.wav
+    ts = int(time.time() * 1000)
+    spell_slug = re.sub(r"[^a-zA-Z0-9]+", "_", spell).strip("_").lower()
+    fname = f"{spell_slug}_{username}_{ts}.wav"
+    out_path = os.path.join(OUT_DIR, fname)
+
+    sf.write(out_path, audio, TARGET_SR, subtype="PCM_16")
+    return out_path
+
+
+def submit_recordings(
+    username: str,
+    lumos_path: Optional[str],
+    nox_path: Optional[str],
+    alohomora_path: Optional[str],
+    wingardium_path: Optional[str],
+    accio_path: Optional[str],
+    reparo_path: Optional[str],
+) -> str:
+    user = sanitize_username(username)
+
+    pairs: List[Tuple[str, Optional[str]]] = [
+        ("Lumos", lumos_path),
+        ("Nox", nox_path),
+        ("Alohomora", alohomora_path),
+        ("Wingardium Leviosa", wingardium_path),
+        ("Accio", accio_path),
+        ("Reparo", reparo_path),
+    ]
+
+    saved = []
+    skipped = []
+
+    for spell, path in pairs:
+        out = save_one_from_path(path, spell, user)
+        if out:
+            saved.append(f"{spell} -> {os.path.basename(out)}")
+        else:
+            skipped.append(spell)
+
+    lines = []
+    if saved:
+        lines.append("Saved recordings:")
+        lines += [f"- {s}" for s in saved]
+    if skipped:
+        lines.append("")
+        lines.append("Missing (not provided):")
+        lines += [f"- {s}" for s in skipped]
+    if not lines:
+        return "No audio captured. Please record at least one spell."
+
+    return "\n".join(lines)
+
+
+def build_ui() -> gr.Blocks:
+    with gr.Blocks(title="Spell Recorder") as demo:
+        gr.Markdown("""
+        # Spell Recorder
+        Record any of the listed spells and press Submit. You can use your microphone directly (preferred) or upload a file.
+
+        Spells to collect: Lumos, Nox, Alohomora, Wingardium Leviosa, Accio, Reparo.
+        """)
+
+        with gr.Row():
+            username = gr.Textbox(label="Your Name (for filename)", placeholder="e.g., harry_p" , autofocus=True)
+
+        with gr.Row():
+            with gr.Column():
+                lumos = gr.Audio(label="Lumos", sources=["microphone", "upload"], type="filepath")
+                nox = gr.Audio(label="Nox", sources=["microphone", "upload"], type="filepath")
+                alohomora = gr.Audio(label="Alohomora", sources=["microphone", "upload"], type="filepath")
+            with gr.Column():
+                wingardium = gr.Audio(label="Wingardium Leviosa", sources=["microphone", "upload"], type="filepath")
+                accio = gr.Audio(label="Accio", sources=["microphone", "upload"], type="filepath")
+                reparo = gr.Audio(label="Reparo", sources=["microphone", "upload"], type="filepath")
+
+        submit = gr.Button("Submit")
+        result = gr.Markdown()
+
+        submit.click(
+            fn=submit_recordings,
+            inputs=[username, lumos, nox, alohomora, wingardium, accio, reparo],
+            outputs=[result],
+        )
+
+        gr.Markdown("""
+        Notes:
+        - Files are saved in the app's `recordings/` folder using: `<spell>_<username>_<timestamp>.wav`.
+        - 16 kHz mono WAV is used to make model training consistent.
+        - You don't have to record all spells at once—submit whatever you have.
+        """)
+
+    return demo
+
+
+demo = build_ui()
+
+if __name__ == "__main__":
+    # Works locally and on Hugging Face Spaces
+    demo.launch()

requirements.txt

@@ -0,0 +1,4 @@
+gradio
+numpy
+soundfile
+scipy