| |
| """profile_clips.py - per-second signal profile of the public sample clips on tinydeskoutstanding.video. |
| For every whole second of every clip: mean luma, mean saturation, motion (mean absolute |
| difference between consecutive frames), audio RMS in dBFS, plus clip-level stream facts and |
| where the reference soundtrack lines up. Writes JSON Lines to stdout. |
| Usage: python3 profile_clips.py labels.json samples_dir reference_file > out.jsonl""" |
| import hashlib, json, os, subprocess, sys |
| import numpy as np |
| sys.path.insert(0, os.path.dirname(os.path.abspath(__file__))) |
| from soundtrack_match import load, onset_envelope, best_offset |
|
|
| W, H = 72, 128 |
| BASE = "https://tinydeskoutstanding.video" |
|
|
| def stream_facts(path): |
| out = subprocess.run(["ffprobe", "-v", "error", "-show_streams", "-of", "json", path], |
| capture_output=True, text=True, check=True).stdout |
| st = json.loads(out)["streams"] |
| v = next(s for s in st if s["codec_type"] == "video") |
| a = next(s for s in st if s["codec_type"] == "audio") |
| num, den = map(int, v["avg_frame_rate"].split("/")) |
| return {"width": v["width"], "height": v["height"], "fps": num / den, |
| "video_duration_s": round(float(v["duration"]), 3), |
| "audio_duration_s": round(float(a["duration"]), 3), |
| "audio_sample_rate": int(a["sample_rate"])} |
|
|
| def frames(path): |
| raw = subprocess.run(["ffmpeg", "-v", "error", "-i", path, "-vf", f"scale={W}:{H}", |
| "-f", "rawvideo", "-pix_fmt", "rgb24", "-"], |
| capture_output=True, check=True).stdout |
| return np.frombuffer(raw, np.uint8).reshape(-1, H, W, 3).astype(np.float32) / 255.0 |
|
|
| def per_second(path, fps): |
| f = frames(path) |
| luma = (0.2126 * f[..., 0] + 0.7152 * f[..., 1] + 0.0722 * f[..., 2]).mean(axis=(1, 2)) |
| mx, mn = f.max(axis=3), f.min(axis=3) |
| sat = np.where(mx > 0, (mx - mn) / np.maximum(mx, 1e-6), 0).mean(axis=(1, 2)) |
| motion = np.concatenate([[0.0], np.abs(np.diff(f, axis=0)).mean(axis=(1, 2, 3))]) |
| audio = load(path) |
| rows, fps_i = [], round(fps) |
| for sec in range(int(len(f) // fps_i)): |
| sl = slice(sec * fps_i, (sec + 1) * fps_i) |
| a = audio[sec * 8000:(sec + 1) * 8000] |
| rms = float(np.sqrt((a ** 2).mean())) if len(a) else 0.0 |
| rows.append({"second": sec, |
| "luma_mean": round(float(luma[sl].mean()), 4), |
| "saturation_mean": round(float(sat[sl].mean()), 4), |
| "motion_mean": round(float(motion[sl].mean()), 4), |
| "motion_max": round(float(motion[sl].max()), 4), |
| "audio_rms_dbfs": round(20 * np.log10(rms + 1e-9), 1)}) |
| return rows |
|
|
| if __name__ == "__main__": |
| labels, folder, ref_path = json.load(open(sys.argv[1])), sys.argv[2], sys.argv[3] |
| ref_env = onset_envelope(load(ref_path)) |
| measured_at = os.environ.get("MEASURED_AT", "") |
| for src, label in labels.items(): |
| path = os.path.join(folder, os.path.basename(src)) |
| sha = hashlib.sha256(open(path, "rb").read()).hexdigest() |
| facts = stream_facts(path) |
| lag, peak = best_offset(ref_env, onset_envelope(load(path))) |
| clip = {"clip_id": sha[:12], "sha256": sha, "source_url": BASE + src, |
| "site_label": label, "is_reference": os.path.abspath(path) == os.path.abspath(ref_path), |
| **facts, "soundtrack_offset_s": round(lag, 2), "soundtrack_match": round(peak, 2), |
| "measured_at": measured_at} |
| for r in per_second(path, facts["fps"]): |
| print(json.dumps({**clip, **r}, ensure_ascii=False)) |
|
|
