fix(worker): align gpu-direct title defaults (size/margin/bold) with CPU vfb path
CI/CD Pipeline / Dedup Check - skip PR tests when covered by push pipeline (pull_request) Successful in 2s
CI/CD Pipeline / Check if frontend-only change (pull_request) Successful in 2s
CI/CD Pipeline / Check push changed paths (pull_request) Has been skipped
CI/CD Pipeline / Frontend Lint (pull_request) Has been skipped
CI/CD Pipeline / Frontend Unit Tests (pull_request) Has been skipped
Preview Deploy / Deploy Preview Environment (pull_request) Successful in 2m12s
CI/CD Pipeline / PR Build Web Image (pull_request) Has been skipped
PR Automation / Auto Approve on CI Green (pull_request) Successful in 3m6s
CI/CD Pipeline / Build Staging API Image (pull_request) Has been skipped
CI/CD Pipeline / Build Staging Web Image (pull_request) Has been skipped
CI/CD Pipeline / Build Staging Worker Image (pull_request) Has been skipped
CI/CD Pipeline / Retag skipped Staging API Image (pull_request) Has been skipped
CI/CD Pipeline / Retag skipped Staging Web Image (pull_request) Has been skipped
CI/CD Pipeline / Retag skipped Staging Worker Image (pull_request) Has been skipped
CI/CD Pipeline / Deploy Staging (Watchtower auto-deploy) (pull_request) Has been skipped
CI/CD Pipeline / Staging E2E Tests (pull_request) Has been skipped
CI/CD Pipeline / Staging API Integration Tests (pull_request) Has been skipped
CI/CD Pipeline / ACR Image Cleanup (pull_request) Has been skipped
CI/CD Pipeline / PR Build Worker Image (pull_request) Successful in 2m26s
AI Code Review / AI Code Review (pull_request) Successful in 6m55s
CI/CD Pipeline / Unit Tests (pull_request) Successful in 9m28s
CI/CD Pipeline / Integration Tests (pull_request) Successful in 9m32s
CI/CD Pipeline / PR Build API Image (pull_request) Failing after 8m40s
CI/CD Pipeline / Validate - Style (pull_request) Successful in 11m15s
CI/CD Pipeline / Validate - Python (mypy + alembic) (pull_request) Successful in 11m45s
PR Automation / Auto Merge on CI Green + Approved (pull_request) Successful in 10m46s
CI/CD Pipeline / Validate - Security (pull_request) Has been cancelled
CI/CD Pipeline / Build Production API Image (pull_request) Has been cancelled
CI/CD Pipeline / Build Production Web Image (pull_request) Has been cancelled
CI/CD Pipeline / Build Production Worker Image (pull_request) Has been cancelled
CI/CD Pipeline / Deploy Production (pull_request) Has been cancelled
CI/CD Pipeline / Production Browser E2E (pull_request) Has been cancelled
CI/CD Pipeline / Canary Release to Production (pull_request) Has been cancelled
CI/CD Pipeline / CI Gate (pull_request) Has been cancelled

PR #2095 fixed title baseline positioning and width-scaling but didn't fully
align default constants with the CPU video_filter_builder path, causing GPU
rendered titles to appear slightly smaller / higher / bolder-differently than
the CPU/ASS preview the template was authored against. This shows up in both
single and batch GPU renders since every variant shares the same gpu_direct
pipeline.

Root causes in PR #2095 defaults:
- Default title size 36@720p vs config_schemas DEFAULT 48 / vfb default 48
- Top/bottom margin 40@720p (PAD16+margin24) vs vfb _scale_title_len(50)
- Subtitle bottom margin 60@720p vs vfb 50
- Faux bold used same-color 1px border (white-on-white invisible for default
  white text; red-on-red for colored) vs vfb black 2px (intentional choice
  per #2001 to avoid double-print/halo artifact)
- margin_top was treated as absolute y offset, now correctly added on top of
  base margin (matches vfb additive semantics)
- subtitle stroke/bold block referenced uninitialized s_borderw (ruff F821)
  — added proper init + stroke parsing consistent with title block

Changes:
- TITLE_DEFAULT_MARGIN_TOP/BOTTOM = 50 (was 24+PAD=40)
- SUBTITLE_DEFAULT_MARGIN_BOTTOM = 50 (was 60)
- TITLE_FAUX_BOLD_WIDTH = 2, border color = #000000 (was 1, same as text)
- default title size 48@720p (was 36)
- margin_top from cfg added on top of base 50 (additive, same as vfb)
- subtitle: init s_borderw/s_border_color, parse stroke dict before bold check
- bottom position: y = h - th - margin_bottom (correct baseline; previously
  used margin_top variable which was misleading but numerically equivalent
  before margin split; now uses explicit margin_bottom)

Tests updated to new expected scaled values (1280x720 scale=1.778: fontsize 85,
y=89, bold borderw=4 black; 1080x1920 vertical: fontsize 72, y=75;
margin_top=100 user offset => y=267). Added test_title_bold_false_disables_faux_bold.

Batch investigation notes (separate findings, not bugs in this PR):
- per-variant plan config is correctly deep-copied via clone_plan_for_variant
  (config=dict(source.config or {})); title text per-variant via titles[]
  override; voice per-variant independent download with #1749 strict guards;
  bgm merged via merge_bgm_config — batch passthrough chain is correct.
- Worker generation concurrency = 2 (compose.yml default); USER_PENDING_LIMIT=3,
  6 tasks will enqueue in two waves (first 3 → then next 3 as workers free up);
  GLOBAL_PENDING_LIMIT allows it. This is expected behavior, not a bug.
- Legacy key mismatch: subtitle_render_engine.py reads plan_config['subtitle_config']
  (snake_case) but normalize_plan_config writes 'subtitle' short key. This file
  is not imported by unified_render_service (which uses render_subtitles.py with
  explicit subtitle_config= parameter), so no runtime impact; left for separate
  cleanup.
- Frontend usePlanConfigLoader.ts reads config.title_config snake_case while
  backend writes 'title' short key — frontend-only issue for subsequent edit
  sessions, out of backend scope.
This commit is contained in:
saas-backend-agent
2026-09-29 18:08:18 +08:00
parent 3be06c5763
commit dfc5e5a5b6
2 changed files with 79 additions and 53 deletions
@@ -30,11 +30,16 @@ EDGE_CROP_MAX_PCT = 0.05
# 均以 720p 为基准(见前端 titleCanvas.ts 注释 scale=videoWidth/720,types.ts "px @720p"),
# 非 720p 输出时按 video_width / TITLE_SIZE_REF_WIDTH 等比缩放,保证成片位置与前端预览一致。
TITLE_SIZE_REF_WIDTH = 720
# 与前端 titleCanvas.ts 对齐:top 时文本 top-edge 距视频顶 = PAD(16@720p) + margin_top(默认 24@720p)
TITLE_PAD_TOP = 16
TITLE_DEFAULT_MARGIN_TOP = 24
TITLE_DEFAULT_MARGIN_BOTTOM = 24
SUBTITLE_DEFAULT_MARGIN_BOTTOM = 60 # 字幕距底边距(720p 基准,与前端字幕面板默认对齐)
# 与 video_filter_builder.build_title_drawtext_filter(CPU 路径)和 ass_subtitle_builder 对齐:
# - top/bottom 默认 margin 50@720p(vfb 用 _scale_title_len(50, w),即 y=50 / y=h-th-50)
# - margin_top 字段:前端编辑器 marginTop 滑块,叠加在默认 margin 之上(#2095 支持)
# - PAD 概念仅用于前端 Canvas 预览;ffmpeg drawtext y 是 baseline,无 font metrics 可用,
# 直接用统一 50@720p baseline 位置即可保持三端(GPU/CPU/前端视觉)一致。
TITLE_DEFAULT_MARGIN_TOP = 50 # top 位置 baseline 默认距顶 50@720p(与 vfb/CPU 路径一致)
TITLE_DEFAULT_MARGIN_BOTTOM = 50 # bottom 位置 baseline 默认距底 50@720p
SUBTITLE_DEFAULT_MARGIN_BOTTOM = 50 # 字幕距底边距 50@720p(与 vfb 一致)
TITLE_MARGIN_TOP_FROM_CFG_DEFAULT = 24 # 前端 marginTop 滑块默认值(用户未传时叠加 0)
TITLE_FAUX_BOLD_WIDTH = 2 # 仿粗黑色描边宽度(与 vfb 一致,2@720p 黑色细描边)
def _scale_title_len(value, video_width: int):
@@ -507,7 +512,7 @@ def build_direct_render(
except (TypeError, ValueError):
t_size_720 = 0
if t_size_720 <= 0:
t_size_720 = 36 # 与 config_schemas 默认 48 接近;AI Avatar 前端默认 48,剪辑前端默认 28,取中
t_size_720 = 48 # 与 config_schemas.DEFAULT_EDIT_PLAN_CONFIG.title.size=48 及 vfb 默认一致
t_size = _scale_title_len(t_size_720, output_width)
# stroke/shadow 长度字段也需 720p→输出分辨率缩放
t_color = str(t_cfg.get("color", t_cfg.get("font_color", "#ffffff")))
@@ -525,14 +530,22 @@ def build_direct_render(
# baseline 需再下移约 0.85*fontsize;但 drawtext 表达式无法引用 fontsize 变量,
# 这里直接用 (PAD + margin_top)@720p 缩放后作为 y(即让 baseline≈顶部内边距位置),
# 实际中文字符会自然向下延伸,视觉位置与前端预览(textBaseline=middle 居中到 firstLineY)一致。
_t_margin_top_raw = t_cfg.get("margin_top", TITLE_DEFAULT_MARGIN_TOP)
# margin_top:前端滑块值(默认 24@720p),叠加在默认 50@720p 基线之上
_t_user_margin_top = t_cfg.get("margin_top")
try:
_t_margin_top_720 = int(_t_margin_top_raw)
_t_user_margin_top_720 = int(_t_user_margin_top) if _t_user_margin_top is not None else 0
except (TypeError, ValueError):
_t_margin_top_720 = TITLE_DEFAULT_MARGIN_TOP
t_margin_top = _scale_title_len(TITLE_PAD_TOP + _t_margin_top_720, output_width)
# bottom margin(标题放在 bottom 时也支持)
t_margin_bottom = _scale_title_len(TITLE_PAD_TOP + TITLE_DEFAULT_MARGIN_BOTTOM, output_width)
_t_user_margin_top_720 = 0
t_margin_top_720 = TITLE_DEFAULT_MARGIN_TOP + _t_user_margin_top_720
t_margin_top = _scale_title_len(t_margin_top_720, output_width)
# bottom margin(标题放在 bottom 时):用户 margin_bottom 透传,默认 50@720p
_t_user_margin_bottom = t_cfg.get("margin_bottom")
try:
_t_user_margin_bottom_720 = int(_t_user_margin_bottom) if _t_user_margin_bottom is not None else 0
except (TypeError, ValueError):
_t_user_margin_bottom_720 = 0
t_margin_bottom_720 = TITLE_DEFAULT_MARGIN_BOTTOM + _t_user_margin_bottom_720
t_margin_bottom = _scale_title_len(t_margin_bottom_720, output_width)
t_borderw = 0
t_border_color = "#000000"
t_box = False
@@ -566,11 +579,11 @@ def build_direct_render(
t_shadow_x = _scale_title_len(t_shadow_x_720, output_width)
t_shadow_y = _scale_title_len(t_shadow_y_720, output_width)
# bold/italic:drawtext 原生无粗斜体选项;通过同色描边模拟粗体
t_bold = bool(t_cfg.get("bold", True)) # 与 ASS 路径默认 bold=True 对齐
t_bold = bool(t_cfg.get("bold", True)) # 与 ASS/vfb 路径默认 bold=True 对齐
if t_bold and t_borderw < 1:
# 同色描边宽度 1@720p,按宽度缩放保证各分辨率视觉一致
t_borderw = max(1, _scale_title_len(1, output_width))
t_border_color = t_color # 用文字色描边模拟加粗
# 粗体未配用户描边时:黑色细描边 2@720p(与 vfb 一致,避免同色描边导致重影)
t_borderw = _scale_title_len(TITLE_FAUX_BOLD_WIDTH, output_width)
t_border_color = "#000000" # 黑色细描边模拟粗体
# ── 解析 subtitle_config ──
s_cfg = dict(subtitle_config) if isinstance(subtitle_config, dict) else {}
@@ -595,8 +608,21 @@ def build_direct_render(
s_pos_x, s_pos_y = None, None
s_margin_top = _scale_title_len(60, output_width) # subtitle top (not commonly used)
s_margin_bottom = _scale_title_len(SUBTITLE_DEFAULT_MARGIN_BOTTOM, output_width)
s_borderw = max(1, _scale_title_len(2, output_width)) # 字幕默认描边保证可读性
# subtitle stroke/bold:先解析用户 stroke,再按 bold 默认补描边
s_borderw = 0
s_border_color = "#000000"
_s_stroke = s_cfg.get("stroke")
if isinstance(_s_stroke, dict) and _s_stroke.get("enabled", False):
try:
s_borderw = _scale_title_len(int(float(_s_stroke.get("width", 2))), output_width)
except (TypeError, ValueError):
s_borderw = 0
s_border_color = str(_s_stroke.get("color", "#000000"))
s_bold = bool(s_cfg.get("bold", False))
if s_bold and s_borderw < 1:
# 粗体默认黑色细描边 2@720p(与 title/CPU vfb 一致)
s_borderw = _scale_title_len(TITLE_FAUX_BOLD_WIDTH, output_width)
s_border_color = "#000000"
# 静态字幕:static_subtitle_text 非空时构造全片长 segment(0 → total_duration)
static_text = (static_subtitle_text or "").strip()
+36 -36
View File
@@ -85,7 +85,7 @@ def _patch_pipeline_helpers(monkeypatch):
# ---------------------------------------------------------------------------
class TestNoConfigBackwardCompat:
def test_default_title_drawtext_white_top(self, monkeypatch):
"""不传 title_config 时:白字、top、36@720 按 width 缩放到 64、y=71(PAD16+margin24)。"""
"""不传 title_config 时:白字、top、48@720 按 width 缩放到 85、y=89(50@720p baseline)。"""
import video_processing.gpu_direct_pipeline as gdp
_patch_pipeline_helpers(monkeypatch)
@@ -101,15 +101,15 @@ class TestNoConfigBackwardCompat:
)
fc = " ".join(plan.filter_complex)
assert "fontcolor=0xffffff" in fc
# 默认 position=top → y=71(scale(16+24)=71),不含 h-th
assert "y=71" in fc
# 默认 position=top → y=89(scale(50)=89,与 CPU/vfb 一致),不含 h-th
assert "y=89" in fc
assert "h-th" not in fc
# 默认字号 36@720 经 1280/720 缩放 = 64
assert "fontsize=64" in fc
# 默认字号 48@720 经 1280/720 缩放 = 85
assert "fontsize=85" in fc
assert "text='默认标题'" in fc
# 默认粗体:drawtext 无原生粗体,用同色描边 borderw=1@720(1280/720→scale=2)模拟
assert "borderw=2" in fc
assert "bordercolor=0xffffff" in fc
# 默认粗体:黑色细描边 borderw=2@720(scale=4),与 CPU vfb 一致避免重影
assert "borderw=4" in fc
assert "bordercolor=0x000000" in fc
def test_no_subtitle_when_none(self, monkeypatch):
import video_processing.gpu_direct_pipeline as gdp
@@ -153,8 +153,8 @@ class TestTitleStylePassthrough:
assert "fontcolor=0xff0000" in fc
# 60@720 按 1280/720 缩放 = 107
assert "fontsize=107" in fc
# top 位置 y=71(scale(PAD16+margin_top24)=71)
assert "y=71" in fc
# top 位置默认 margin 50@720p → scale=89(未传 margin_top 用默认)
assert "y=89" in fc
assert "h-th" not in fc
def test_title_position_center(self, monkeypatch):
@@ -218,9 +218,9 @@ class TestTitleStylePassthrough:
# 阴影层 + 主字层 = 2 条 drawtext
assert drawtext_count == 2
joined = " ".join(plan.filter_complex)
# shadow offset 3@720 经 1280/720 缩放 = 5;默认 position=top → y=71
# shadow offset 3@720 经 1280/720 缩放 = 5;默认 position=top → y=89
assert "x=(w-text_w)/2+5" in joined
assert "y=71+5" in joined
assert "y=89+5" in joined
assert "h-th" not in joined
def test_title_font_override(self, monkeypatch):
@@ -282,10 +282,10 @@ class TestTitleStylePassthrough:
assert "(w-text_w)*0.3000" in fc
assert "(h-text_h)*0.6000" in fc
assert "h-th-" not in fc
assert "y=71" not in fc
assert "y=89" not in fc
def test_title_margin_top_respected(self, monkeypatch):
"""margin_top 透传:100@720 → scale(16+100)=206@1280。"""
"""margin_top 透传:100@720,叠加默认 50 → 150@720 → scale=267@1280。"""
import video_processing.gpu_direct_pipeline as gdp
_patch_pipeline_helpers(monkeypatch)
@@ -300,8 +300,8 @@ class TestTitleStylePassthrough:
title_config={"text": "远离顶部", "position": "top", "margin_top": 100},
)
fc = " ".join(plan.filter_complex)
# scale(PAD16 + margin_top100) = 116*1280/720 = 206
assert "y=206" in fc
# 默认 50 + margin_top100 = 150@720 → scale=267
assert "y=267" in fc
def test_title_default_bold_true(self, monkeypatch):
"""不传 bold 时默认粗体:drawtext 用同色描边 borderw 模拟。"""
@@ -319,9 +319,9 @@ class TestTitleStylePassthrough:
title_config={"text": "粗体"},
)
fc = " ".join(plan.filter_complex)
# bold=True 默认:同色描边 width=1@720 → scale=2,颜色与文字相同(白色)
assert "borderw=2" in fc
assert "bordercolor=0xffffff" in fc
# bold=True 默认:黑色细描边 width=2@720 → scale=4,与 CPU vfb 一致
assert "borderw=4" in fc
assert "bordercolor=0x000000" in fc
assert "text='粗体'" in fc
def test_title_bold_false_disables_faux_bold(self, monkeypatch):
@@ -359,8 +359,8 @@ class TestTitleStylePassthrough:
title_config={"text": "底部", "position": "bottom"},
)
fc = " ".join(plan.filter_complex)
# bottom margin PAD16+24=40 → scale=71
assert "y=h-th-71" in fc
# bottom margin 50@720 → scale=89
assert "y=h-th-89" in fc
def test_title_size_scales_by_width_not_height(self, monkeypatch):
"""不同分辨率下同 @720 基准的 size 等比缩放:1080x1920 下 36→54。"""
@@ -378,10 +378,10 @@ class TestTitleStylePassthrough:
clip_volumes=[1.0],
)
fc = " ".join(plan.filter_complex)
# default size 36@720 → 1080w = 54
assert "fontsize=54" in fc
# top margin 40@720 → 60
assert "y=60" in fc
# default size 48@720 → 1080w = 72
assert "fontsize=72" in fc
# top margin 50@720 → 75
assert "y=75" in fc
# ---------------------------------------------------------------------------
@@ -434,8 +434,8 @@ class TestSubtitlePassthrough:
assert "fontcolor=0x0000ff" in fc
# sub size 28@720 经 1280/720 缩放 = 50
assert "fontsize=50" in fc
# bottom margin 60@720 → 107
assert "y=h-th-107" in fc
# bottom margin 50@720 → 89
assert "y=h-th-89" in fc
def test_subtitle_disabled_hides_subs(self, monkeypatch):
import video_processing.gpu_direct_pipeline as gdp
@@ -644,18 +644,18 @@ class TestCombinedConfig:
bgm_config={"volume": 0.2, "fade_in": 0.5, "fade_out": 1.0},
)
fc = " ".join(plan.filter_complex)
# 标题:size 50@720 → 89,top margin 40@720 → 71,stroke 2@720 → 4
# 标题:size 50@720 → 89,top margin 50@720 → 89,stroke 2@720 → 4
assert "text='主标题'" in fc
assert "fontcolor=0xffff00" in fc
assert "fontsize=89" in fc
assert "y=71" in fc
assert "y=89" in fc
assert "borderw=4" in fc
# 字幕:size 24@720 → 43,bottom margin 60@720 → 107
# 字幕:size 24@720 → 43,bottom margin 50@720 → 89(默认不加粗)
assert "text='成片全字幕'" in fc
assert "fontsize=43" in fc
assert "y=h-th-107" in fc
assert "y=h-th-89" in fc
assert "fontcolor=0xffffff" in fc
# subtitle 默认 bold=True 也会加同色描边(width=1→2);但若主标题已有 stroke 不影响字幕独立 borderw=4(s 默认 2 像素@720→scale=4)
# subtitle 默认 bold=False,无额外描边(用户未开 stroke)
# BGM
assert "volume=0.200" in fc
assert "afade=t=in:st=0:d=0.50" in fc
@@ -688,7 +688,7 @@ class TestEdgeCases:
assert "drawtext=" in fc
def test_bold_title_increases_borderw(self, monkeypatch):
"""bold=True 时若原无描边,自动加 borderw=1 用同色描边模拟加粗。"""
"""bold=True 时若原无描边,自动加 borderw 黑色细描边模拟加粗(与 CPU vfb 一致,避免重影)。"""
import video_processing.gpu_direct_pipeline as gdp
_patch_pipeline_helpers(monkeypatch)
@@ -703,9 +703,9 @@ class TestEdgeCases:
title_config={"text": "粗体", "bold": True, "color": "#ff0000"},
)
fc = " ".join(plan.filter_complex)
# 加粗模拟 borderw>=1,且 bordercolor 跟字体色一致(0xff0000)
assert "borderw=" in fc
assert "bordercolor=0xff0000" in fc
# 仿粗用黑色细描边 2@720→scale=4,不跟文字色(避免同色描边重影)
assert "borderw=4" in fc
assert "bordercolor=0x000000" in fc
def test_asr_and_static_subtitle_asr_wins(self, monkeypatch):
"""同时传 static_subtitle_text 和 ASR segments 时,ASR 优先(不插入静态全文)。"""