scripts/31_reconstruct_media.py

#!/usr/bin/env python3
"""Reconstruct manifest-backed media without using hidden or server-absolute row paths."""
from __future__ import annotations

import argparse
import hashlib
import io
import json
import os
import subprocess
import tarfile
import urllib.request
from pathlib import Path

from PIL import Image, ImageDraw, ImageFont

ROOT = Path(os.environ.get("STREAMING_OMNI_ROOT", "/home/XuWenqi/ZhouZhifan/StreamingOmniDatasetsWork"))
PIPER = ROOT / ".cache/tools/piper/piper"
PIPER_MODEL = ROOT / ".cache/tools/piper/en_US-lessac-medium.onnx"


def sha256(path: Path):
    digest = hashlib.sha256()
    with path.open("rb") as handle:
        for block in iter(lambda: handle.read(1024 * 1024), b""):
            digest.update(block)
    return digest.hexdigest()


def generate_text_image(item: dict, target: Path):
    width, height = item.get("width", 640), item.get("height", 480)
    image = Image.new("RGB", (width, height), "white")
    draw = ImageDraw.Draw(image)
    font = ImageFont.load_default()
    text = str(item["text"])
    words, lines, current = text.split(), [], ""
    for word in words:
        candidate = (current + " " + word).strip()
        if draw.textlength(candidate, font=font) > width - 40 and current:
            lines.append(current)
            current = word
        else:
            current = candidate
    if current:
        lines.append(current)
    draw.multiline_text((20, 20), "\n".join(lines), fill="black", font=font, spacing=8)
    target.parent.mkdir(parents=True, exist_ok=True)
    image.save(target, format="PNG", optimize=False)


def generate_speech(item: dict, target: Path):
    if not PIPER.exists() or not PIPER_MODEL.exists():
        raise RuntimeError("Piper executable/model missing under project-local .cache/tools/piper")
    target.parent.mkdir(parents=True, exist_ok=True)
    subprocess.run(
        [str(PIPER), "--model", str(PIPER_MODEL), "--output_file", str(target)],
        input=str(item["text"]), text=True, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
    )


def download_member(item: dict, target: Path, cache: Path):
    url, _, member = str(item["source_url"]).partition("#")
    cache.mkdir(parents=True, exist_ok=True)
    archive = cache / (hashlib.sha256(url.encode()).hexdigest() + ".tar")
    if not archive.exists():
        with urllib.request.urlopen(url, timeout=120) as response, archive.open("wb") as output:
            while block := response.read(1024 * 1024):
                output.write(block)
    extracted = cache / (hashlib.sha256((url + "#" + member).encode()).hexdigest() + ".mp4")
    if not extracted.exists():
        with tarfile.open(archive) as tar:
            candidates = [entry for entry in tar.getmembers() if entry.name == member or entry.name.endswith("/" + member)]
            if not candidates:
                raise RuntimeError(f"member not found in upstream archive: {member}")
            source = tar.extractfile(candidates[0])
            if source is None:
                raise RuntimeError(f"member is not a regular file: {member}")
            extracted.write_bytes(source.read())
    target.parent.mkdir(parents=True, exist_ok=True)
    start, end = float(item.get("start_sec") or 0), float(item.get("end_sec") or 0)
    command = ["ffmpeg", "-y", "-hide_banner", "-loglevel", "error"]
    if start > 0:
        command += ["-ss", str(start)]
    command += ["-i", str(extracted)]
    if end > start:
        command += ["-t", str(end - start)]
    command += ["-c", "copy", str(target)]
    subprocess.run(command, check=True)


def main() -> int:
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("--config", required=True)
    parser.add_argument("--output-dir", type=Path, required=True)
    parser.add_argument("--limit", type=int)
    parser.add_argument("--offset", type=int, default=0)
    args = parser.parse_args()
    manifest = ROOT / "manifests" / f"{args.config}.jsonl"
    items = [json.loads(line) for line in manifest.read_text().splitlines() if line.strip()]
    items = items[args.offset:args.offset + args.limit if args.limit else None]
    results = []
    for item in items:
        target = args.output_dir / item["output_relative_path"]
        try:
            if item["generator_family"] == "pil_text_renderer":
                generate_text_image(item, target)
            elif item["generator_family"] == "piper_tts":
                generate_speech(item, target)
            elif item["reconstruction_method"] in {"clip", "download"}:
                download_member(item, target, ROOT / ".cache/reconstruction")
            else:
                raise RuntimeError("unsupported reconstruction recipe")
            results.append({"manifest_id": item["manifest_id"], "success": True, "path": str(target), "sha256": sha256(target)})
        except Exception as exc:
            results.append({"manifest_id": item["manifest_id"], "success": False, "error": str(exc)[:500]})
    print(json.dumps(results, indent=2))
    return 0 if results and all(item["success"] for item in results) else 1


if __name__ == "__main__":
    raise SystemExit(main())