dsh-drawai 0.4.2 → 0.4.4

package/lib/motion/shotcraft/upstream/assets/scripts/sfx-source.py

@@ -1,423 +1,423 @@
-#!/usr/bin/env python3
-"""Find Freesound SFX and freeze licensed MP3 previews in a video project.
-
-Python standard library; get/analyze also require ffmpeg and ffprobe.
-No API key, global cache, login, or proxy configuration changes.
-"""
-
-import argparse
-import array
-from concurrent.futures import ThreadPoolExecutor
-from datetime import datetime, timezone
-import hashlib
-from html.parser import HTMLParser
-import json
-import math
-from pathlib import Path
-import re
-import shutil
-import subprocess
-import sys
-import tempfile
-from urllib.parse import unquote, urlencode, urljoin, urlparse
-from urllib.request import Request, urlopen
-
-
-SITE = "https://freesound.org"
-MANIFEST = "audio-manifest.json"
-CREDITS = "FREESOUND-CREDITS.md"
-MAX_BYTES = 32 * 1024 * 1024
-VOID = set("area base br col embed hr img input link meta param source track wbr".split())
-
-
-class Node:
-    def __init__(self, tag="root", attrs=()):
-        self.tag, self.attrs, self.children = tag, dict(attrs), []
-
-    def nodes(self):
-        yield self
-        for child in self.children:
-            if isinstance(child, Node):
-                yield from child.nodes()
-
-    def text(self):
-        return " ".join(c.text() if isinstance(c, Node) else c for c in self.children).strip()
-
-    def has_class(self, name):
-        return name in self.attrs.get("class", "").split()
-
-
-class Document(HTMLParser):
-    def __init__(self, content):
-        super().__init__(convert_charrefs=True)
-        self.root = Node()
-        self.stack = [self.root]
-        self.feed(content)
-        self.close()
-
-    def handle_starttag(self, tag, attrs):
-        node = Node(tag, attrs)
-        self.stack[-1].children.append(node)
-        if tag not in VOID:
-            self.stack.append(node)
-
-    def handle_startendtag(self, tag, attrs):
-        self.handle_starttag(tag, attrs)
-        if tag not in VOID:
-            self.handle_endtag(tag)
-
-    def handle_endtag(self, tag):
-        for i in range(len(self.stack) - 1, 0, -1):
-            if self.stack[i].tag == tag:
-                del self.stack[i:]
-                break
-
-    def handle_data(self, data):
-        self.stack[-1].children.append(data)
-
-
-def request(url):
-    req = Request(url, headers={"User-Agent": "video-shotcraft/1.0 (Freesound SFX sourcing)"})
-    with urlopen(req, timeout=30) as response:
-        data = response.read(MAX_BYTES + 1)
-        if len(data) > MAX_BYTES:
-            raise ValueError("Response exceeds 32 MiB; select a shorter sound or download the original manually")
-        return data, response.geturl(), response.headers.get_content_type()
-
-
-def web_page(url):
-    data, final_url, content_type = request(url)
-    if urlparse(final_url).hostname != "freesound.org" or content_type != "text/html":
-        raise ValueError("Expected a Freesound HTML page")
-    return Document(data.decode("utf-8")), final_url
-
-
-def sound_id(value):
-    value = value.removeprefix("fs:")
-    if not re.fullmatch(r"[1-9][0-9]*", value):
-        raise ValueError(f"Invalid Freesound ID: {value!r}; use fs:12345")
-    return value
-
-
-def license_code(url):
-    parsed = urlparse(url)
-    if parsed.scheme not in {"http", "https"} or parsed.hostname not in {"creativecommons.org", "www.creativecommons.org"}:
-        return None
-    path = parsed.path.rstrip("/")
-    if path == "/publicdomain/zero/1.0":
-        return "CC0-1.0"
-    match = re.fullmatch(r"/licenses/by/(1\.0|2\.0|2\.5|3\.0|4\.0)", path)
-    return f"CC-BY-{match[1]}" if match else None
-
-
-def check_license(code, policy):
-    if policy == "any":
-        return
-    if code == "CC0-1.0" or (policy == "by" and code and code.startswith("CC-BY-")):
-        return
-    raise ValueError(f"License {code or 'unknown'} is outside --license {policy}; CC BY requires --license by")
-
-
-def player_data(node, page_author=None):
-    attrs = node.attrs
-    sid = sound_id(attrs["data-sound-id"])
-    title, author = attrs.get("data-title"), page_author or attrs.get("data-username")
-    duration = float(attrs["data-duration"])
-    if not title or not author or not math.isfinite(duration) or duration <= 0:
-        raise ValueError(f"Incomplete Freesound metadata for fs:{sid}")
-    preview = attrs.get("data-mp3", "")
-    parsed = urlparse(preview)
-    if (parsed.scheme != "https" or parsed.hostname != "cdn.freesound.org"
-            or not re.fullmatch(rf"/previews/[0-9]+/{sid}_[0-9]+-(lq|hq)\.mp3", parsed.path)):
-        raise ValueError(f"Missing or unexpected preview URL for fs:{sid}")
-    return {"id": f"fs:{sid}", "title": title, "author": author,
-            "duration_seconds": duration, "preview_url": preview.replace("-lq.mp3", "-hq.mp3"),
-            "downloads": int(attrs.get("data-num-downloads") or 0)}
-
-
-def stats(nodes):
-    rating, count = None, None
-    for node in nodes:
-        label = node.attrs.get("aria-label", "")
-        if label.startswith("Average rating of "):
-            rating = float(label.removeprefix("Average rating of "))
-        if node.has_class("bw-rating__count"):
-            found = re.search(r"\(([0-9]+)(?: ratings?)?\)", node.text())
-            if found:
-                count = int(found[1])
-    return {"rating": rating, "rating_count": count}
-
-
-def parse_search(doc):
-    rows = []
-    for block in doc.root.nodes():
-        if not block.has_class("bw-search__result"):
-            continue
-        nodes = list(block.nodes())
-        player = next((n for n in nodes if n.attrs.get("data-sound-id")), None)
-        if player is None:
-            raise ValueError("Freesound search markup changed: result has no sound metadata")
-        row = player_data(player)
-        links = [urljoin(SITE, n.attrs.get("href", "")) for n in nodes if n.tag == "a"]
-        row["page_url"] = next((u for u in links if u.endswith(f"/sounds/{sound_id(row['id'])}/")), None)
-        row["license_label"] = next((n.attrs["title"][9:] for n in nodes
-                                     if n.attrs.get("title", "").startswith("License: ")), "unknown")
-        desc = next((n for n in nodes if any(c.startswith("max-h-60px")
-                     for c in n.attrs.get("class", "").split())), None)
-        row["description"] = " ".join((desc.attrs.get("title") or desc.text()).split()) if desc else ""
-        row.update(stats(nodes))
-        rows.append(row)
-    if not rows and not re.search(r"no (?:sounds|results)|0 sounds|couldn't find|did not match", doc.root.text(), re.I):
-        raise ValueError("No recognizable search results; check Freesound availability or update the parser")
-    return rows
-
-
-def search(args):
-    if args.license == "any":
-        labels = None
-        filters = []
-    else:
-        labels = {"Creative Commons 0"} | ({"Attribution"} if args.license == "by" else set())
-        filters = ['license:("Creative Commons 0" OR "Attribution")' if args.license == "by"
-                   else 'license:"Creative Commons 0"']
-    filters.append(f"duration:[{args.min_duration} TO {args.max_duration if args.max_duration is not None else '*'}]")
-    found = {}
-    for page in range(1, args.pages + 1):
-        query = urlencode({"q": args.query, "f": " ".join(filters), "page": page, "s": "Automatic by relevance"})
-        doc, _ = web_page(f"{SITE}/search/?{query}")
-        for row in parse_search(doc):
-            # Enforce locally too; never trust a server-side filter alone.
-            if labels is not None and row["license_label"] not in labels:
-                continue
-            if row["duration_seconds"] < args.min_duration:
-                continue
-            if args.max_duration is not None and row["duration_seconds"] > args.max_duration:
-                continue
-            found.setdefault(row["id"], row)
-    result = {"provider": "freesound", "query": args.query, "license_policy": args.license,
-              "ordering": "relevance; popularity is supporting evidence only",
-              "results": list(found.values())[:args.limit]}
-    emit(result, args.output)
-    return 0
-
-
-def detail(sid):
-    doc, final_url = web_page(f"{SITE}/s/{sid}/")
-    nodes = list(doc.root.nodes())
-    player = next((n for n in nodes if n.attrs.get("data-sound-id") == sid and n.attrs.get("data-mp3")), None)
-    canonical = re.fullmatch(rf"/people/([^/]+)/sounds/{sid}/", urlparse(final_url).path)
-    if player is None or canonical is None:
-        raise ValueError(f"Missing canonical sound page for fs:{sid}")
-    # Large players omit data-username; the canonical page supplies the creator.
-    row = player_data(player, unquote(canonical[1]))
-    row["page_url"] = final_url
-    licenses = {n.attrs["href"] for n in nodes if n.tag == "a"
-                and urlparse(n.attrs.get("href", "")).hostname in {"creativecommons.org", "www.creativecommons.org"}}
-    if len(licenses) != 1:
-        raise ValueError(f"Missing or ambiguous license on fs:{sid}")
-    row["license_url"] = licenses.pop()
-    row["license"] = license_code(row["license_url"])
-    row["description"] = next((n.attrs.get("content", "") for n in nodes
-                               if n.attrs.get("name") == "twitter:description"), "")
-    row.update(stats(nodes))
-    return row
-
-
-def require_audio_tools():
-    if not all(shutil.which(name) for name in ("ffmpeg", "ffprobe")):
-        raise ValueError("get/analyze require ffmpeg and ffprobe on PATH")
-
-
-def analyze(path):
-    require_audio_tools()
-    path = Path(path).resolve()
-    probe = subprocess.run(["ffprobe", "-v", "error", "-select_streams", "a:0", "-show_entries",
-                            "stream=codec_name,sample_rate,channels:format=duration", "-of", "json", str(path)],
-                           capture_output=True, text=True, check=True)
-    info = json.loads(probe.stdout)
-    stream = info["streams"][0]
-    channels = int(stream["channels"])
-    if not 1 <= channels <= 8:
-        raise ValueError("Expected an audio file with 1–8 channels")
-    rate, window = 48000, 0.01
-    block_size = int(rate * window) * channels * 4
-    energies, peak, squares, samples = [], 0.0, 0.0, 0
-    # Preserve channels: downmixing first can hide opposite-polarity peaks.
-    with tempfile.TemporaryFile() as errors:
-        with subprocess.Popen(["ffmpeg", "-v", "error", "-i", str(path), "-map", "0:a:0", "-ar", str(rate),
-                               "-f", "f32le", "-acodec", "pcm_f32le", "pipe:1"],
-                              stdout=subprocess.PIPE, stderr=errors) as process:
-            while True:
-                raw = process.stdout.read(block_size)
-                if not raw:
-                    break
-                values = array.array("f")
-                values.frombytes(raw)
-                if sys.byteorder != "little":
-                    values.byteswap()
-                energy = sum(v * v for v in values)
-                energies.append(math.sqrt(energy / len(values)))
-                peak = max(peak, max(abs(v) for v in values))
-                squares += energy
-                samples += len(values)
-            if process.wait() != 0:
-                errors.seek(0)
-                raise ValueError(errors.read().decode("utf-8", errors="replace"))
-    if not samples or not peak:
-        raise ValueError("Audio is empty or silent")
-    loudest = max(energies)
-    active = [i for i, e in enumerate(energies) if e >= loudest * 0.01]
-    duration = samples / (rate * channels)
-    return {"codec": stream["codec_name"], "sample_rate": int(stream["sample_rate"]), "channels": channels,
-            "duration_seconds": round(duration, 6), "sample_peak_dbfs": round(20 * math.log10(peak), 3),
-            "rms_dbfs": round(20 * math.log10(math.sqrt(squares / samples)), 3),
-            "energy_window_seconds": window,
-            "energy_peak_seconds": round(min(duration, (energies.index(loudest) + 0.5) * window), 6),
-            "active_start_seconds": round(active[0] * window, 6),
-            "active_end_seconds": round(min(duration, (active[-1] + 1) * window), 6),
-            "activity_threshold_db_below_peak_rms": -40}
-
-
-def write_json(path, data):
-    path = Path(path)
-    path.parent.mkdir(parents=True, exist_ok=True)
-    with tempfile.NamedTemporaryFile(mode="w", encoding="utf-8", dir=path.parent, delete=False) as tmp:
-        json.dump(data, tmp, ensure_ascii=False, indent=2, allow_nan=False)
-        tmp.write("\n")
-    Path(tmp.name).replace(path)
-
-
-def emit(data, output=None):
-    if output:
-        write_json(output, data)
-    print(json.dumps(data, ensure_ascii=False, indent=2, allow_nan=False))
-
-
-def digest(path):
-    return hashlib.sha256(Path(path).read_bytes()).hexdigest()
-
-
-def write_credits(out, rows):
-    # Plain text inside a fence keeps uploaded titles from becoming Markdown links/instructions.
-    lines = ["# Freesound audio credits", "", "Generated from audio-manifest.json; preserve with the video.", "",
-             "CC BY entries require attribution in the published video's credits/description.", "",
-             "Downloaded files are lossy MP3 previews, not uploader originals.", ""]
-    for row in rows:
-        lines += ["~~~text", f"{row['file']} — {row['title']} by {row['author']}".replace("~", "\\~"),
-                  row["page_url"], f"{row['license']}: {row['license_url']}",
-                  "Edits: " + ("; ".join(row.get("changes", [])) or "None recorded; update if edited."), "~~~", ""]
-    (out / CREDITS).write_text("\n".join(lines), encoding="utf-8")
-
-
-def download_one(sid, out, policy, existing):
-    row = detail(sid)  # Refresh the actual license, even when search/cache metadata is stale.
-    check_license(row["license"], policy)
-    dst = out / f"fs-{sid}.mp3"
-    old = next((r for r in existing if r["id"] == row["id"]), None)
-    if dst.exists():
-        if old is None or old.get("file") != dst.name or digest(dst) != old.get("sha256"):
-            raise ValueError(f"Refusing to overwrite untracked/modified file: {dst}")
-        return {**old, **row, "analysis": analyze(dst), "license_checked_at": datetime.now(timezone.utc).isoformat()}
-    payload, final_url, mime = request(row["preview_url"])
-    if urlparse(final_url).hostname != "cdn.freesound.org" or mime not in {"audio/mpeg", "audio/mp3", "application/octet-stream"}:
-        raise ValueError("Expected a Freesound MP3 preview, not an HTML error page")
-    with tempfile.NamedTemporaryFile(dir=out, suffix=".mp3", delete=False) as tmp:
-        tmp.write(payload)
-        tmp_path = Path(tmp.name)
-    try:
-        measurements = analyze(tmp_path)  # Decode successfully before admitting the file.
-        if measurements["codec"] != "mp3":
-            raise ValueError("Preview is not MP3 audio")
-        row.update({"provider": "freesound", "file": dst.name, "sha256": digest(tmp_path),
-                    "download_url": final_url, "download_kind": "mp3-preview-hq", "is_preview": True,
-                    "downloaded_at": datetime.now(timezone.utc).isoformat(),
-                    "license_checked_at": datetime.now(timezone.utc).isoformat(), "changes": [], "analysis": measurements})
-        # Exclusive creation: another process/user's file is never replaced.
-        with dst.open("xb") as target:
-            target.write(payload)
-        return row
-    finally:
-        tmp_path.unlink(missing_ok=True)
-
-
-def get(args):
-    require_audio_tools()
-    ids = list(dict.fromkeys(sound_id(value) for value in args.ids))
-    out = Path(args.out_dir).resolve()
-    out.mkdir(parents=True, exist_ok=True)
-    lock = out / ".freesound-download.lock"
-    try:
-        handle = lock.open("x")
-    except FileExistsError:
-        raise ValueError(f"Download already active in {out}; remove {lock.name} only if a previous run crashed")
-    try:
-        with handle:
-            return get_locked(args, ids, out)
-    finally:
-        lock.unlink(missing_ok=True)
-
-
-def get_locked(args, ids, out):
-    manifest_path = out / MANIFEST
-    manifest = json.loads(manifest_path.read_text(encoding="utf-8")) if manifest_path.exists() else {"version": 1, "assets": []}
-    if manifest.get("version") != 1 or not isinstance(manifest.get("assets"), list):
-        raise ValueError("Unsupported audio manifest; use a separate output directory")
-    existing = manifest["assets"]
-
-    def run(sid):
-        try:
-            return {"ok": True, "asset": download_one(sid, out, args.license, existing)}
-        except (OSError, ValueError, KeyError, IndexError, subprocess.SubprocessError) as error:
-            return {"ok": False, "id": f"fs:{sid}", "error": str(error)}
-
-    results = []
-    with ThreadPoolExecutor(max_workers=args.workers) as pool:
-        for result in pool.map(run, ids):
-            results.append(result)
-            if result["ok"]:
-                row = result["asset"]
-                manifest["assets"] = [r for r in manifest["assets"] if r["id"] != row["id"]] + [row]
-                write_json(manifest_path, manifest)
-                write_credits(out, manifest["assets"])
-    emit({"out_dir": str(out), "results": results})
-    return 0 if all(r["ok"] for r in results) else 1
-
-
-def main(argv=None):
-    parser = argparse.ArgumentParser(description=__doc__)
-    commands = parser.add_subparsers(dest="command", required=True)
-    find = commands.add_parser("search", help="Search public Freesound pages; emit JSON candidates")
-    find.add_argument("query")
-    find.add_argument("--license", choices=("cc0", "by", "any"), default="any", help="any does not filter by license; by permits CC0 + CC BY")
-    find.add_argument("--min-duration", type=float, default=0)
-    find.add_argument("--max-duration", type=float)
-    find.add_argument("--limit", type=int, default=8)
-    find.add_argument("--pages", type=int, choices=range(1, 4), default=1)
-    find.add_argument("--output", help="Optional project-local JSON path")
-    fetch = commands.add_parser("get", help="Download IDs, verify licenses/audio, save manifest and credits")
-    fetch.add_argument("ids", nargs="+")
-    fetch.add_argument("--out-dir", required=True, help="Video project's public/audio/sfx directory")
-    fetch.add_argument("--license", choices=("cc0", "by", "any"), default="any")
-    fetch.add_argument("--workers", type=int, choices=range(1, 5), default=3)
-    measure = commands.add_parser("analyze", help="Measure decoded audio (10 ms RMS peak; not LUFS/true peak)")
-    measure.add_argument("file")
-    measure.add_argument("--output")
-    args = parser.parse_args(argv)
-    try:
-        if args.command == "search":
-            if not 1 <= args.limit <= 45 or not math.isfinite(args.min_duration) or args.min_duration < 0:
-                raise ValueError("Use --limit 1–45 and a nonnegative finite --min-duration")
-            if args.max_duration is not None and (not math.isfinite(args.max_duration) or args.max_duration <= args.min_duration):
-                raise ValueError("--max-duration must be finite and greater than --min-duration")
-            return search(args)
-        if args.command == "get":
-            return get(args)
-        emit(analyze(args.file), args.output)
-        return 0
-    except (OSError, ValueError, KeyError, IndexError, subprocess.SubprocessError) as error:
-        print(f"sfx-source: {error}", file=sys.stderr)
-        return 1
-
-
-if __name__ == "__main__":
-    sys.exit(main())
+#!/usr/bin/env python3
+"""Find Freesound SFX and freeze licensed MP3 previews in a video project.
+
+Python standard library; get/analyze also require ffmpeg and ffprobe.
+No API key, global cache, login, or proxy configuration changes.
+"""
+
+import argparse
+import array
+from concurrent.futures import ThreadPoolExecutor
+from datetime import datetime, timezone
+import hashlib
+from html.parser import HTMLParser
+import json
+import math
+from pathlib import Path
+import re
+import shutil
+import subprocess
+import sys
+import tempfile
+from urllib.parse import unquote, urlencode, urljoin, urlparse
+from urllib.request import Request, urlopen
+
+
+SITE = "https://freesound.org"
+MANIFEST = "audio-manifest.json"
+CREDITS = "FREESOUND-CREDITS.md"
+MAX_BYTES = 32 * 1024 * 1024
+VOID = set("area base br col embed hr img input link meta param source track wbr".split())
+
+
+class Node:
+    def __init__(self, tag="root", attrs=()):
+        self.tag, self.attrs, self.children = tag, dict(attrs), []
+
+    def nodes(self):
+        yield self
+        for child in self.children:
+            if isinstance(child, Node):
+                yield from child.nodes()
+
+    def text(self):
+        return " ".join(c.text() if isinstance(c, Node) else c for c in self.children).strip()
+
+    def has_class(self, name):
+        return name in self.attrs.get("class", "").split()
+
+
+class Document(HTMLParser):
+    def __init__(self, content):
+        super().__init__(convert_charrefs=True)
+        self.root = Node()
+        self.stack = [self.root]
+        self.feed(content)
+        self.close()
+
+    def handle_starttag(self, tag, attrs):
+        node = Node(tag, attrs)
+        self.stack[-1].children.append(node)
+        if tag not in VOID:
+            self.stack.append(node)
+
+    def handle_startendtag(self, tag, attrs):
+        self.handle_starttag(tag, attrs)
+        if tag not in VOID:
+            self.handle_endtag(tag)
+
+    def handle_endtag(self, tag):
+        for i in range(len(self.stack) - 1, 0, -1):
+            if self.stack[i].tag == tag:
+                del self.stack[i:]
+                break
+
+    def handle_data(self, data):
+        self.stack[-1].children.append(data)
+
+
+def request(url):
+    req = Request(url, headers={"User-Agent": "video-shotcraft/1.0 (Freesound SFX sourcing)"})
+    with urlopen(req, timeout=30) as response:
+        data = response.read(MAX_BYTES + 1)
+        if len(data) > MAX_BYTES:
+            raise ValueError("Response exceeds 32 MiB; select a shorter sound or download the original manually")
+        return data, response.geturl(), response.headers.get_content_type()
+
+
+def web_page(url):
+    data, final_url, content_type = request(url)
+    if urlparse(final_url).hostname != "freesound.org" or content_type != "text/html":
+        raise ValueError("Expected a Freesound HTML page")
+    return Document(data.decode("utf-8")), final_url
+
+
+def sound_id(value):
+    value = value.removeprefix("fs:")
+    if not re.fullmatch(r"[1-9][0-9]*", value):
+        raise ValueError(f"Invalid Freesound ID: {value!r}; use fs:12345")
+    return value
+
+
+def license_code(url):
+    parsed = urlparse(url)
+    if parsed.scheme not in {"http", "https"} or parsed.hostname not in {"creativecommons.org", "www.creativecommons.org"}:
+        return None
+    path = parsed.path.rstrip("/")
+    if path == "/publicdomain/zero/1.0":
+        return "CC0-1.0"
+    match = re.fullmatch(r"/licenses/by/(1\.0|2\.0|2\.5|3\.0|4\.0)", path)
+    return f"CC-BY-{match[1]}" if match else None
+
+
+def check_license(code, policy):
+    if policy == "any":
+        return
+    if code == "CC0-1.0" or (policy == "by" and code and code.startswith("CC-BY-")):
+        return
+    raise ValueError(f"License {code or 'unknown'} is outside --license {policy}; CC BY requires --license by")
+
+
+def player_data(node, page_author=None):
+    attrs = node.attrs
+    sid = sound_id(attrs["data-sound-id"])
+    title, author = attrs.get("data-title"), page_author or attrs.get("data-username")
+    duration = float(attrs["data-duration"])
+    if not title or not author or not math.isfinite(duration) or duration <= 0:
+        raise ValueError(f"Incomplete Freesound metadata for fs:{sid}")
+    preview = attrs.get("data-mp3", "")
+    parsed = urlparse(preview)
+    if (parsed.scheme != "https" or parsed.hostname != "cdn.freesound.org"
+            or not re.fullmatch(rf"/previews/[0-9]+/{sid}_[0-9]+-(lq|hq)\.mp3", parsed.path)):
+        raise ValueError(f"Missing or unexpected preview URL for fs:{sid}")
+    return {"id": f"fs:{sid}", "title": title, "author": author,
+            "duration_seconds": duration, "preview_url": preview.replace("-lq.mp3", "-hq.mp3"),
+            "downloads": int(attrs.get("data-num-downloads") or 0)}
+
+
+def stats(nodes):
+    rating, count = None, None
+    for node in nodes:
+        label = node.attrs.get("aria-label", "")
+        if label.startswith("Average rating of "):
+            rating = float(label.removeprefix("Average rating of "))
+        if node.has_class("bw-rating__count"):
+            found = re.search(r"\(([0-9]+)(?: ratings?)?\)", node.text())
+            if found:
+                count = int(found[1])
+    return {"rating": rating, "rating_count": count}
+
+
+def parse_search(doc):
+    rows = []
+    for block in doc.root.nodes():
+        if not block.has_class("bw-search__result"):
+            continue
+        nodes = list(block.nodes())
+        player = next((n for n in nodes if n.attrs.get("data-sound-id")), None)
+        if player is None:
+            raise ValueError("Freesound search markup changed: result has no sound metadata")
+        row = player_data(player)
+        links = [urljoin(SITE, n.attrs.get("href", "")) for n in nodes if n.tag == "a"]
+        row["page_url"] = next((u for u in links if u.endswith(f"/sounds/{sound_id(row['id'])}/")), None)
+        row["license_label"] = next((n.attrs["title"][9:] for n in nodes
+                                     if n.attrs.get("title", "").startswith("License: ")), "unknown")
+        desc = next((n for n in nodes if any(c.startswith("max-h-60px")
+                     for c in n.attrs.get("class", "").split())), None)
+        row["description"] = " ".join((desc.attrs.get("title") or desc.text()).split()) if desc else ""
+        row.update(stats(nodes))
+        rows.append(row)
+    if not rows and not re.search(r"no (?:sounds|results)|0 sounds|couldn't find|did not match", doc.root.text(), re.I):
+        raise ValueError("No recognizable search results; check Freesound availability or update the parser")
+    return rows
+
+
+def search(args):
+    if args.license == "any":
+        labels = None
+        filters = []
+    else:
+        labels = {"Creative Commons 0"} | ({"Attribution"} if args.license == "by" else set())
+        filters = ['license:("Creative Commons 0" OR "Attribution")' if args.license == "by"
+                   else 'license:"Creative Commons 0"']
+    filters.append(f"duration:[{args.min_duration} TO {args.max_duration if args.max_duration is not None else '*'}]")
+    found = {}
+    for page in range(1, args.pages + 1):
+        query = urlencode({"q": args.query, "f": " ".join(filters), "page": page, "s": "Automatic by relevance"})
+        doc, _ = web_page(f"{SITE}/search/?{query}")
+        for row in parse_search(doc):
+            # Enforce locally too; never trust a server-side filter alone.
+            if labels is not None and row["license_label"] not in labels:
+                continue
+            if row["duration_seconds"] < args.min_duration:
+                continue
+            if args.max_duration is not None and row["duration_seconds"] > args.max_duration:
+                continue
+            found.setdefault(row["id"], row)
+    result = {"provider": "freesound", "query": args.query, "license_policy": args.license,
+              "ordering": "relevance; popularity is supporting evidence only",
+              "results": list(found.values())[:args.limit]}
+    emit(result, args.output)
+    return 0
+
+
+def detail(sid):
+    doc, final_url = web_page(f"{SITE}/s/{sid}/")
+    nodes = list(doc.root.nodes())
+    player = next((n for n in nodes if n.attrs.get("data-sound-id") == sid and n.attrs.get("data-mp3")), None)
+    canonical = re.fullmatch(rf"/people/([^/]+)/sounds/{sid}/", urlparse(final_url).path)
+    if player is None or canonical is None:
+        raise ValueError(f"Missing canonical sound page for fs:{sid}")
+    # Large players omit data-username; the canonical page supplies the creator.
+    row = player_data(player, unquote(canonical[1]))
+    row["page_url"] = final_url
+    licenses = {n.attrs["href"] for n in nodes if n.tag == "a"
+                and urlparse(n.attrs.get("href", "")).hostname in {"creativecommons.org", "www.creativecommons.org"}}
+    if len(licenses) != 1:
+        raise ValueError(f"Missing or ambiguous license on fs:{sid}")
+    row["license_url"] = licenses.pop()
+    row["license"] = license_code(row["license_url"])
+    row["description"] = next((n.attrs.get("content", "") for n in nodes
+                               if n.attrs.get("name") == "twitter:description"), "")
+    row.update(stats(nodes))
+    return row
+
+
+def require_audio_tools():
+    if not all(shutil.which(name) for name in ("ffmpeg", "ffprobe")):
+        raise ValueError("get/analyze require ffmpeg and ffprobe on PATH")
+
+
+def analyze(path):
+    require_audio_tools()
+    path = Path(path).resolve()
+    probe = subprocess.run(["ffprobe", "-v", "error", "-select_streams", "a:0", "-show_entries",
+                            "stream=codec_name,sample_rate,channels:format=duration", "-of", "json", str(path)],
+                           capture_output=True, text=True, check=True)
+    info = json.loads(probe.stdout)
+    stream = info["streams"][0]
+    channels = int(stream["channels"])
+    if not 1 <= channels <= 8:
+        raise ValueError("Expected an audio file with 1–8 channels")
+    rate, window = 48000, 0.01
+    block_size = int(rate * window) * channels * 4
+    energies, peak, squares, samples = [], 0.0, 0.0, 0
+    # Preserve channels: downmixing first can hide opposite-polarity peaks.
+    with tempfile.TemporaryFile() as errors:
+        with subprocess.Popen(["ffmpeg", "-v", "error", "-i", str(path), "-map", "0:a:0", "-ar", str(rate),
+                               "-f", "f32le", "-acodec", "pcm_f32le", "pipe:1"],
+                              stdout=subprocess.PIPE, stderr=errors) as process:
+            while True:
+                raw = process.stdout.read(block_size)
+                if not raw:
+                    break
+                values = array.array("f")
+                values.frombytes(raw)
+                if sys.byteorder != "little":
+                    values.byteswap()
+                energy = sum(v * v for v in values)
+                energies.append(math.sqrt(energy / len(values)))
+                peak = max(peak, max(abs(v) for v in values))
+                squares += energy
+                samples += len(values)
+            if process.wait() != 0:
+                errors.seek(0)
+                raise ValueError(errors.read().decode("utf-8", errors="replace"))
+    if not samples or not peak:
+        raise ValueError("Audio is empty or silent")
+    loudest = max(energies)
+    active = [i for i, e in enumerate(energies) if e >= loudest * 0.01]
+    duration = samples / (rate * channels)
+    return {"codec": stream["codec_name"], "sample_rate": int(stream["sample_rate"]), "channels": channels,
+            "duration_seconds": round(duration, 6), "sample_peak_dbfs": round(20 * math.log10(peak), 3),
+            "rms_dbfs": round(20 * math.log10(math.sqrt(squares / samples)), 3),
+            "energy_window_seconds": window,
+            "energy_peak_seconds": round(min(duration, (energies.index(loudest) + 0.5) * window), 6),
+            "active_start_seconds": round(active[0] * window, 6),
+            "active_end_seconds": round(min(duration, (active[-1] + 1) * window), 6),
+            "activity_threshold_db_below_peak_rms": -40}
+
+
+def write_json(path, data):
+    path = Path(path)
+    path.parent.mkdir(parents=True, exist_ok=True)
+    with tempfile.NamedTemporaryFile(mode="w", encoding="utf-8", dir=path.parent, delete=False) as tmp:
+        json.dump(data, tmp, ensure_ascii=False, indent=2, allow_nan=False)
+        tmp.write("\n")
+    Path(tmp.name).replace(path)
+
+
+def emit(data, output=None):
+    if output:
+        write_json(output, data)
+    print(json.dumps(data, ensure_ascii=False, indent=2, allow_nan=False))
+
+
+def digest(path):
+    return hashlib.sha256(Path(path).read_bytes()).hexdigest()
+
+
+def write_credits(out, rows):
+    # Plain text inside a fence keeps uploaded titles from becoming Markdown links/instructions.
+    lines = ["# Freesound audio credits", "", "Generated from audio-manifest.json; preserve with the video.", "",
+             "CC BY entries require attribution in the published video's credits/description.", "",
+             "Downloaded files are lossy MP3 previews, not uploader originals.", ""]
+    for row in rows:
+        lines += ["~~~text", f"{row['file']} — {row['title']} by {row['author']}".replace("~", "\\~"),
+                  row["page_url"], f"{row['license']}: {row['license_url']}",
+                  "Edits: " + ("; ".join(row.get("changes", [])) or "None recorded; update if edited."), "~~~", ""]
+    (out / CREDITS).write_text("\n".join(lines), encoding="utf-8")
+
+
+def download_one(sid, out, policy, existing):
+    row = detail(sid)  # Refresh the actual license, even when search/cache metadata is stale.
+    check_license(row["license"], policy)
+    dst = out / f"fs-{sid}.mp3"
+    old = next((r for r in existing if r["id"] == row["id"]), None)
+    if dst.exists():
+        if old is None or old.get("file") != dst.name or digest(dst) != old.get("sha256"):
+            raise ValueError(f"Refusing to overwrite untracked/modified file: {dst}")
+        return {**old, **row, "analysis": analyze(dst), "license_checked_at": datetime.now(timezone.utc).isoformat()}
+    payload, final_url, mime = request(row["preview_url"])
+    if urlparse(final_url).hostname != "cdn.freesound.org" or mime not in {"audio/mpeg", "audio/mp3", "application/octet-stream"}:
+        raise ValueError("Expected a Freesound MP3 preview, not an HTML error page")
+    with tempfile.NamedTemporaryFile(dir=out, suffix=".mp3", delete=False) as tmp:
+        tmp.write(payload)
+        tmp_path = Path(tmp.name)
+    try:
+        measurements = analyze(tmp_path)  # Decode successfully before admitting the file.
+        if measurements["codec"] != "mp3":
+            raise ValueError("Preview is not MP3 audio")
+        row.update({"provider": "freesound", "file": dst.name, "sha256": digest(tmp_path),
+                    "download_url": final_url, "download_kind": "mp3-preview-hq", "is_preview": True,
+                    "downloaded_at": datetime.now(timezone.utc).isoformat(),
+                    "license_checked_at": datetime.now(timezone.utc).isoformat(), "changes": [], "analysis": measurements})
+        # Exclusive creation: another process/user's file is never replaced.
+        with dst.open("xb") as target:
+            target.write(payload)
+        return row
+    finally:
+        tmp_path.unlink(missing_ok=True)
+
+
+def get(args):
+    require_audio_tools()
+    ids = list(dict.fromkeys(sound_id(value) for value in args.ids))
+    out = Path(args.out_dir).resolve()
+    out.mkdir(parents=True, exist_ok=True)
+    lock = out / ".freesound-download.lock"
+    try:
+        handle = lock.open("x")
+    except FileExistsError:
+        raise ValueError(f"Download already active in {out}; remove {lock.name} only if a previous run crashed")
+    try:
+        with handle:
+            return get_locked(args, ids, out)
+    finally:
+        lock.unlink(missing_ok=True)
+
+
+def get_locked(args, ids, out):
+    manifest_path = out / MANIFEST
+    manifest = json.loads(manifest_path.read_text(encoding="utf-8")) if manifest_path.exists() else {"version": 1, "assets": []}
+    if manifest.get("version") != 1 or not isinstance(manifest.get("assets"), list):
+        raise ValueError("Unsupported audio manifest; use a separate output directory")
+    existing = manifest["assets"]
+
+    def run(sid):
+        try:
+            return {"ok": True, "asset": download_one(sid, out, args.license, existing)}
+        except (OSError, ValueError, KeyError, IndexError, subprocess.SubprocessError) as error:
+            return {"ok": False, "id": f"fs:{sid}", "error": str(error)}
+
+    results = []
+    with ThreadPoolExecutor(max_workers=args.workers) as pool:
+        for result in pool.map(run, ids):
+            results.append(result)
+            if result["ok"]:
+                row = result["asset"]
+                manifest["assets"] = [r for r in manifest["assets"] if r["id"] != row["id"]] + [row]
+                write_json(manifest_path, manifest)
+                write_credits(out, manifest["assets"])
+    emit({"out_dir": str(out), "results": results})
+    return 0 if all(r["ok"] for r in results) else 1
+
+
+def main(argv=None):
+    parser = argparse.ArgumentParser(description=__doc__)
+    commands = parser.add_subparsers(dest="command", required=True)
+    find = commands.add_parser("search", help="Search public Freesound pages; emit JSON candidates")
+    find.add_argument("query")
+    find.add_argument("--license", choices=("cc0", "by", "any"), default="any", help="any does not filter by license; by permits CC0 + CC BY")
+    find.add_argument("--min-duration", type=float, default=0)
+    find.add_argument("--max-duration", type=float)
+    find.add_argument("--limit", type=int, default=8)
+    find.add_argument("--pages", type=int, choices=range(1, 4), default=1)
+    find.add_argument("--output", help="Optional project-local JSON path")
+    fetch = commands.add_parser("get", help="Download IDs, verify licenses/audio, save manifest and credits")
+    fetch.add_argument("ids", nargs="+")
+    fetch.add_argument("--out-dir", required=True, help="Video project's public/audio/sfx directory")
+    fetch.add_argument("--license", choices=("cc0", "by", "any"), default="any")
+    fetch.add_argument("--workers", type=int, choices=range(1, 5), default=3)
+    measure = commands.add_parser("analyze", help="Measure decoded audio (10 ms RMS peak; not LUFS/true peak)")
+    measure.add_argument("file")
+    measure.add_argument("--output")
+    args = parser.parse_args(argv)
+    try:
+        if args.command == "search":
+            if not 1 <= args.limit <= 45 or not math.isfinite(args.min_duration) or args.min_duration < 0:
+                raise ValueError("Use --limit 1–45 and a nonnegative finite --min-duration")
+            if args.max_duration is not None and (not math.isfinite(args.max_duration) or args.max_duration <= args.min_duration):
+                raise ValueError("--max-duration must be finite and greater than --min-duration")
+            return search(args)
+        if args.command == "get":
+            return get(args)
+        emit(analyze(args.file), args.output)
+        return 0
+    except (OSError, ValueError, KeyError, IndexError, subprocess.SubprocessError) as error:
+        print(f"sfx-source: {error}", file=sys.stderr)
+        return 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())