@oh-story/dsh 0.1.3 → 0.1.5

package/lib/video-recap/skills/video-assemble/scripts/subtitle_core.py

@@ -0,0 +1,376 @@
+"""Subtitle text shaping, timing, and measured-canvas geometry."""
+
+import os
+import re
+
+from lib import CONFIG
+from assemble_constants import (
+    SUBTITLE_STYLE_REF_H,
+    SUBTITLE_STYLE_REF_W,
+    _SUBTITLE_CLOSING_QUOTES,
+    _SUBTITLE_TERMINAL_PUNCTUATION,
+)
+from media import _ratio_to_float
+
+def _seconds_to_srt_time(seconds):
+    """将秒数转为 SRT 时间格式 HH:MM:SS,mmm"""
+    h = int(seconds // 3600)
+    m = int((seconds % 3600) // 60)
+    s = int(seconds % 60)
+    ms = int((seconds % 1) * 1000)
+    return f"{h:02d}:{m:02d}:{s:02d},{ms:03d}"
+
+
+def _seconds_to_ass_time(seconds):
+    """将秒数转为 ASS 时间格式 H:MM:SS.cc"""
+    centiseconds = int(round(float(seconds) * 100))
+    h = centiseconds // 360000
+    centiseconds %= 360000
+    m = centiseconds // 6000
+    centiseconds %= 6000
+    s = centiseconds // 100
+    cs = centiseconds % 100
+    return f"{h}:{m:02d}:{s:02d}.{cs:02d}"
+
+
+def _subtitle_style_config(canvas=None):
+    """Return the internal default burn-in subtitle style.
+
+    When ``canvas`` ({"width","height"}) is given AND the user has not pinned PlayRes via
+    SUBTITLE_PLAY_RES_X/Y, the style is scaled to that canvas: PlayRes is set to the frame
+    dimensions (so libass never stretches glyphs — the old hardcoded 1280x720 squished
+    portrait text), horizontal metrics scale with width, vertical metrics with height, and
+    the font is additionally capped so a full ``max_chars`` line fits the usable width. A
+    16:9 source (or the 1280x720 default) reproduces the legacy values exactly.
+    """
+    style = {
+        "font_name": CONFIG.get("subtitle_font_name", "Arial"),
+        "font_size": CONFIG.get("subtitle_font_size", 42),
+        "primary_color": CONFIG.get("subtitle_primary_color", "&H00FFFFFF"),
+        "outline_color": CONFIG.get("subtitle_outline_color", "&H00000000"),
+        "outline": CONFIG.get("subtitle_outline", 2),
+        "shadow": CONFIG.get("subtitle_shadow", 1),
+        "alignment": CONFIG.get("subtitle_alignment", 2),
+        "margin_l": CONFIG.get("subtitle_margin_l", 40),
+        "margin_r": CONFIG.get("subtitle_margin_r", 40),
+        "margin_v": CONFIG.get("subtitle_margin_v", 30),
+        "max_chars": CONFIG.get("subtitle_max_chars", 20),
+        "play_res_x": CONFIG.get("subtitle_play_res_x", 1280),
+        "play_res_y": CONFIG.get("subtitle_play_res_y", 720),
+    }
+    pinned = "SUBTITLE_PLAY_RES_X" in os.environ or "SUBTITLE_PLAY_RES_Y" in os.environ
+    cw = int((canvas or {}).get("width", 0) or 0)
+    ch = int((canvas or {}).get("height", 0) or 0)
+    if canvas is None or pinned or cw <= 0 or ch <= 0:
+        return style  # legacy / manually-pinned: unchanged
+
+    base_font = float(style["font_size"])
+    kx = cw / float(SUBTITLE_STYLE_REF_W)  # horizontal metrics ∝ width
+    ky = ch / float(SUBTITLE_STYLE_REF_H)  # vertical metrics ∝ height
+    margin_l = round(float(style["margin_l"]) * kx)
+    margin_r = round(float(style["margin_r"]) * kx)
+    margin_v = round(float(style["margin_v"]) * ky)
+    # height-proportional size, then cap so a full line of CJK glyphs (≈1em wide) fits the
+    # usable width — this is what keeps portrait text on-screen instead of overflowing.
+    usable_w = max(1.0, cw - margin_l - margin_r)
+    width_cap = usable_w / max(1, int(style["max_chars"]))
+    font_size = max(1, int(min(base_font * ky, width_cap)))  # floor so a full line never overflows
+    font_scale = font_size / base_font if base_font else 1.0
+    style.update({
+        "font_size": font_size,
+        "outline": max(0, round(float(style["outline"]) * font_scale)),
+        "shadow": max(0, round(float(style["shadow"]) * font_scale)),
+        "margin_l": margin_l,
+        "margin_r": margin_r,
+        "margin_v": margin_v,
+        "play_res_x": cw,
+        "play_res_y": ch,
+    })
+    return style
+
+
+def _validate_measured_subtitle_coordinate_domain(canvas=None):
+    y_top = int(CONFIG.get("subtitle_y_top", -1))
+    y_bot = int(CONFIG.get("subtitle_y_bot", -1))
+    if y_top < 0 and y_bot < 0:
+        return
+    sar_text = str((canvas or {}).get("sample_aspect_ratio") or "1:1")
+    if abs(_ratio_to_float(sar_text, 0.0) - 1.0) >= 1e-9:
+        raise ValueError(
+            f"字幕带坐标仅支持方形像素画布 (SAR 1:1)；当前 SAR={sar_text}"
+        )
+
+
+def _style_for_measured_subtitle_band(style, canvas=None):
+    """Fit the ASS baseline and font into explicit auto-rotated display-frame Y coordinates."""
+    style = dict(style)
+    y_top = int(CONFIG.get("subtitle_y_top", -1))
+    y_bot = int(CONFIG.get("subtitle_y_bot", -1))
+    canvas_h = int((canvas or {}).get("height", 0) or 0)
+    if y_top < 0 and y_bot < 0:
+        return style
+    _validate_measured_subtitle_coordinate_domain(canvas)
+    if canvas_h <= 0:
+        return style
+    if not 0 <= y_top < y_bot <= canvas_h:
+        raise ValueError(
+            f"字幕带坐标无效: top={y_top}, bot={y_bot}, 画布高度={canvas_h}；"
+            "必须满足 0 <= top < bot <= height"
+        )
+    alignment = int(style.get("alignment", 2))
+    if alignment not in {1, 2, 3}:
+        raise ValueError(
+            "measured subtitle coordinates require a bottom-aligned ASS style "
+            f"(SUBTITLE_ALIGNMENT 1/2/3); got {alignment}"
+        )
+    scale_y = float(style["play_res_y"]) / canvas_h
+    style["margin_v"] = max(0, round((canvas_h - y_bot) * scale_y))
+    current_font = max(1, int(style["font_size"]))
+    current_outline = float(style.get("outline", 0) or 0)
+    current_shadow = float(style.get("shadow", 0) or 0)
+    safe_area = _measured_subtitle_safe_area(style, canvas)
+    available_height = int((safe_area or {}).get("height", 0) or 0)
+    fitted_font = current_font
+    for candidate in range(current_font, 7, -1):
+        scale = candidate / current_font
+        outline = max(1 if current_outline > 0 else 0, round(current_outline * scale))
+        shadow = max(0, round(current_shadow * scale))
+        if candidate * 1.25 + outline * 2 + shadow <= available_height + 1e-6:
+            fitted_font = candidate
+            break
+    else:
+        # Keep the renderer's minimum readable size; visual QC will block because it cannot fit.
+        fitted_font = min(current_font, 8)
+    if fitted_font < current_font:
+        scale = fitted_font / current_font
+        style["font_size"] = fitted_font
+        style["outline"] = max(
+            1 if current_outline > 0 else 0, round(current_outline * scale)
+        )
+        style["shadow"] = max(0, round(current_shadow * scale))
+    return style
+
+
+def _measured_subtitle_safe_area(style, canvas=None):
+    """Return the padded measured band in ASS PlayRes coordinates, or None."""
+    y_top = int(CONFIG.get("subtitle_y_top", -1))
+    y_bot = int(CONFIG.get("subtitle_y_bot", -1))
+    canvas_w = int((canvas or {}).get("width", 0) or 0)
+    canvas_h = int((canvas or {}).get("height", 0) or 0)
+    if y_top < 0 and y_bot < 0:
+        return None
+    _validate_measured_subtitle_coordinate_domain(canvas)
+    if canvas_w <= 0 or canvas_h <= 0:
+        return None
+    if not 0 <= y_top < y_bot <= canvas_h:
+        raise ValueError(
+            f"字幕带坐标无效: top={y_top}, bot={y_bot}, 画布高度={canvas_h}；"
+            "必须满足 0 <= top < bot <= height"
+        )
+    padding = max(0, int(CONFIG.get("subtitle_mask_padding", 4) or 0))
+    safe_top = max(0, y_top - padding)
+    # The ASS style remains bottom-anchored at the measured y_bot. Bottom mask padding hides
+    # source glyph edges but is not usable subtitle layout space; only top padding can extend
+    # the line box without moving its baseline below the measured band.
+    safe_bot = y_bot
+    play_x = int(style.get("play_res_x") or canvas_w)
+    play_y = int(style.get("play_res_y") or canvas_h)
+    scale_y = play_y / canvas_h
+    margin_l = int(style.get("margin_l") or 0)
+    margin_r = int(style.get("margin_r") or 0)
+    return {
+        "x": margin_l,
+        "y": round(safe_top * scale_y),
+        "width": max(1, play_x - margin_l - margin_r),
+        "height": max(1, round((safe_bot - safe_top) * scale_y)),
+        "bottom_margin": max(0, round((canvas_h - safe_bot) * scale_y)),
+    }
+
+
+def _subtitle_display_text(text):
+    """Return display-only subtitle text with trailing sentence punctuation removed.
+
+    Narration/TTS source text stays untouched; this is applied only to SRT/ASS cue text.
+    Closing quotes/brackets are preserved, so 「原声台词。」 renders as 「原声台词」.
+    """
+    text = str(text or "").strip()
+    if not text:
+        return ""
+    suffix = ""
+    while text and text[-1] in _SUBTITLE_CLOSING_QUOTES:
+        suffix = text[-1] + suffix
+        text = text[:-1].rstrip()
+    text = text.rstrip(_SUBTITLE_TERMINAL_PUNCTUATION).rstrip()
+    return (text + suffix).strip()
+
+
+def _subtitle_chunk_weight(text):
+    """Weight raw subtitle chunks for timing, independent of display punctuation cleanup."""
+    core = re.sub(r"\s+", "", str(text or ""))
+    return max(1, len(core))
+
+
+def _subtitle_entry_chunks(raw_chunks):
+    """Pair raw chunks used for timing with their final display text.
+
+    Timing remains based on the raw split topology. Terminal punctuation is stripped only
+    on the emitted text, while quote-only suffix chunks are folded into the previous cue
+    so a closing bracket never renders alone.
+    """
+    chunks = [str(c).strip() for c in (raw_chunks or []) if str(c).strip()]
+    out = []
+    for i, chunk in enumerate(chunks):
+        display = _subtitle_display_text(chunk)
+        if not display:
+            continue
+        if all(ch in _SUBTITLE_CLOSING_QUOTES for ch in display):
+            if out:
+                out[-1]["text"] += display
+            continue
+        out.append({"raw": chunk, "text": display})
+    return out
+
+
+def _normalize_subtitle_text(s):
+    """Normalize Chinese em-dashes in burned subtitle text: a run of one-or-more "—" (incl. "——")
+    collapses to a single "，". Then collapse any resulting double commas ("，，"→"，") so the dash
+    swap never leaves a doubled comma. Empty/None passes through as "" unchanged."""
+    text = str(s or "")
+    if not text:
+        return text
+    text = re.sub(r"—+", "，", text)
+    text = re.sub(r"，{2,}", "，", text)
+    return text
+
+
+def _split_subtitle_chunks(text, max_chars):
+    """Split one narration block (often several sentences) into short display chunks.
+
+    A block is synthesized as one continuous TTS utterance for fluent prosody, but showing the
+    whole paragraph as a single subtitle would force a tall multi-line band and lag the picture.
+    So we cut the block at punctuation into clauses, then greedily pack adjacent clauses into
+    chunks of at most `max_chars` — each chunk renders as ONE readable line synced to its slice of
+    the block's audio. Punctuation stays attached here for lossless splitting; the display layer
+    strips terminal sentence marks per subtitle-cue style."""
+    text = str(text).strip()
+    if not text:
+        return []
+    breakers = "，。！？、；：…—,.!?;:"
+    clauses, buf = [], ""
+    for ch in text:
+        buf += ch
+        if ch in breakers:
+            clauses.append(buf)
+            buf = ""
+    if buf.strip():
+        clauses.append(buf)
+    # Any single clause longer than max_chars is hard-wrapped so no chunk ever exceeds one line.
+    # Balance those pieces instead of slicing exactly at max_chars: a 21-character clause must
+    # not become a readable 20-character cue followed by a 1-character flash.
+    sized = []
+    for clause in clauses:
+        if len(clause) <= max_chars:
+            sized.append(clause)
+        else:
+            piece_count = (len(clause) + max_chars - 1) // max_chars
+            base, extra = divmod(len(clause), piece_count)
+            cursor = 0
+            for piece_index in range(piece_count):
+                width = base + (1 if piece_index < extra else 0)
+                sized.append(clause[cursor:cursor + width])
+                cursor += width
+    chunks, cur = [], ""
+    for clause in sized:
+        sentence_closed = cur.rstrip().endswith(tuple(_SUBTITLE_TERMINAL_PUNCTUATION))
+        if cur and (sentence_closed or len(cur) + len(clause) > max_chars):
+            chunks.append(cur)
+            cur = clause
+        else:
+            cur += clause
+    if cur.strip():
+        chunks.append(cur)
+    return [c.strip() for c in chunks if c.strip()]
+
+
+def _subtitle_entries(narration):
+    """Collect subtitle entries from final TTS segment placement.
+
+    Each placed segment is split into short one-line chunks and its played window
+    [actual_place_start, actual_place_end] is distributed across them in proportion to character
+    count — karaoke-style timing that keeps each line on screen only while it is roughly being
+    spoken, instead of holding a whole paragraph for the segment's full duration."""
+    max_chars = int(CONFIG.get("subtitle_max_chars", 20))
+    entries = []
+    for seg in narration:
+        if not isinstance(seg, dict):
+            continue
+        text = str(seg.get("spoken_text") or seg.get("narration", "")).strip()
+        if not text:
+            continue
+        try:
+            start = float(seg.get("actual_place_start", seg["start"]))
+            end = float(seg.get("actual_place_end", seg["end"]))
+        except (KeyError, TypeError, ValueError):
+            continue
+        if end - start < 0.1:
+            continue
+        chunks = _subtitle_entry_chunks(_split_subtitle_chunks(text, max_chars))
+        if not chunks:
+            continue
+        if len(chunks) == 1:
+            entries.append({"start": start, "end": end, "text": chunks[0]["text"]})
+            continue
+        total_chars = sum(_subtitle_chunk_weight(c["raw"]) for c in chunks) or 1
+        span = end - start
+        cursor = start
+        for i, chunk in enumerate(chunks):
+            weight = _subtitle_chunk_weight(chunk["raw"])
+            chunk_end = end if i == len(chunks) - 1 else cursor + span * (weight / total_chars)
+            if i > 0 and chunk_end - cursor < 0.05:
+                # slice too short to show on its own — fold the text into the previous line of THIS
+                # block (i>0) and extend its end, so no chunk is ever silently dropped.
+                entries[-1]["text"] += chunk["text"]
+                entries[-1]["end"] = chunk_end
+            else:
+                entries.append({"start": cursor, "end": chunk_end, "text": chunk["text"]})
+            cursor = chunk_end
+    return entries
+
+
+def _distribute_chunks(chunks, start, end):
+    """Distribute [start,end] across raw chunks while emitting display-clean text.
+
+    Terminal subtitle punctuation is visual-only: it is stripped from final cue text,
+    but the raw split chunks remain the timing topology.
+    """
+    chunks = _subtitle_entry_chunks(chunks)
+    if not chunks or end - start < 0.1:
+        return []
+    if len(chunks) == 1:
+        return [{"start": start, "end": end, "text": chunks[0]["text"]}]
+    total_chars = sum(_subtitle_chunk_weight(c["raw"]) for c in chunks) or 1
+    span = end - start
+    out, cursor = [], start
+    for i, chunk in enumerate(chunks):
+        weight = _subtitle_chunk_weight(chunk["raw"])
+        chunk_end = end if i == len(chunks) - 1 else cursor + span * (weight / total_chars)
+        if out and chunk_end - cursor < 0.05:
+            out[-1]["text"] += chunk["text"]
+            out[-1]["end"] = chunk_end
+        else:
+            out.append({"start": cursor, "end": chunk_end, "text": chunk["text"]})
+        cursor = chunk_end
+    return out
+
+
+def _bracketed_original_chunks(text, start, end, max_chars):
+    """Split original dialogue into timed chunks wrapped in 「」 for visual distinction."""
+    raw = str(text).strip()
+    if raw.startswith("「") and raw.endswith("」"):
+        raw = raw[1:-1].strip()
+    chunks = _split_subtitle_chunks(raw, max_chars)
+    if chunks:
+        chunks = list(chunks)
+        chunks[0] = "「" + chunks[0]
+        chunks[-1] = chunks[-1] + "」"
+    return _distribute_chunks(chunks, start, end)

package/lib/video-recap/skills/video-assemble/scripts/subtitle_render.py

@@ -0,0 +1,88 @@
+"""SRT/ASS serialization for narration and original-dialogue subtitles."""
+
+from source_subtitles import _combined_subtitle_entries
+from subtitle_core import (
+    _normalize_subtitle_text,
+    _seconds_to_ass_time,
+    _seconds_to_srt_time,
+    _style_for_measured_subtitle_band,
+    _subtitle_entries,
+    _subtitle_style_config,
+)
+
+def _generate_srt(narration, work_dir, video_duration=None):
+    """将解说脚本转为 SRT 字幕文件，使用实际音频放置时间。video_duration 给定时，原声留白处补烧原声字幕。"""
+    srt_lines = []
+    entries = (_subtitle_entries(narration) if video_duration is None
+               else _combined_subtitle_entries(narration, work_dir, video_duration))
+    # entries are already split into short one-line chunks, so no wrapping here.
+    for idx, entry in enumerate(entries, start=1):
+        start_ts = _seconds_to_srt_time(entry["start"])
+        end_ts = _seconds_to_srt_time(entry["end"])
+        srt_lines.append(str(idx))
+        srt_lines.append(f"{start_ts} --> {end_ts}")
+        srt_lines.append(_normalize_subtitle_text(entry["text"]))
+        srt_lines.append("")
+    srt_path = work_dir / "subtitles.srt"
+    srt_path.write_text("\n".join(srt_lines), encoding="utf-8")
+    return srt_path
+
+
+def _escape_ass_text(text):
+    """Escape user text for an ASS dialogue Text field."""
+    return (
+        str(text)
+        .replace("\\", "\\\\")
+        .replace("{", "\\{")
+        .replace("}", "\\}")
+        .replace("\r\n", "\n")
+        .replace("\r", "\n")
+        .replace("\n", "\\N")
+    )
+
+
+def _generate_ass(narration, work_dir, video_duration=None, canvas=None):
+    """Generate an ASS subtitle file for readable hard-sub rendering. video_duration given ⇒ also
+    burn the original dialogue (from ASR) during the original-audio gaps. canvas ({"width","height"})
+    scales the style to the real frame so portrait/竖屏 subtitles are not stretched."""
+    style = _style_for_measured_subtitle_band(_subtitle_style_config(canvas), canvas)
+    ass_lines = [
+        "[Script Info]",
+        "ScriptType: v4.00+",
+        "WrapStyle: 0",
+        "ScaledBorderAndShadow: yes",
+        f"PlayResX: {int(style['play_res_x'])}",
+        f"PlayResY: {int(style['play_res_y'])}",
+        "",
+        "[V4+ Styles]",
+        (
+            "Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, "
+            "OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, "
+            "ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, "
+            "Alignment, MarginL, MarginR, MarginV, Encoding"
+        ),
+        (
+            "Style: Default,"
+            f"{style['font_name']},{style['font_size']},{style['primary_color']},&H000000FF,"
+            f"{style['outline_color']},&H64000000,0,0,0,0,100,100,0,0,1,"
+            f"{style['outline']},{style['shadow']},{style['alignment']},"
+            f"{style['margin_l']},{style['margin_r']},{style['margin_v']},1"
+        ),
+        "",
+        "[Events]",
+        "Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text",
+    ]
+    entries = (_subtitle_entries(narration) if video_duration is None
+               else _combined_subtitle_entries(narration, work_dir, video_duration))
+    # entries are already split into short one-line chunks, so no wrapping here.
+    for entry in entries:
+        text = _escape_ass_text(_normalize_subtitle_text(entry["text"]))
+        ass_lines.append(
+            "Dialogue: 0,"
+            f"{_seconds_to_ass_time(entry['start'])},{_seconds_to_ass_time(entry['end'])},"
+            f"Default,,0,0,0,,{text}"
+        )
+
+    ass_path = work_dir / "subtitles.ass"
+    ass_path.write_text("\n".join(ass_lines) + "\n", encoding="utf-8")
+    return ass_path