intent-router | brief.md(主题/关键词,或 viral-chaser 报告) | script/intent.json(档位+主题) | Stage 0 |
reference-concepts | viral-chaser 报告(main 喂入,可选) | reference-driven/concepts.md | Stage 1:只吃报告出概念,不做下载/转写/抽帧;无报告跳过 |
story-develop | intent.json | script/story.md | Stage 2 |
script-write | story.md | script/script.md(含 enhancement_cues + delivery_cues) | Stage 3 |
script-self-eval | script.md | script/self-eval.json(N 维分,任一维 <3 必返工) | Stage 3b |
storyboard-build | script.md | storyboard/storyboard.json | Stage 4 |
shot-decompose | storyboard.json | storyboard/shot_decompose.json(每镜首尾帧+运动+variation_type) | Stage 5 |
character-register | storyboard.json + 人物描述 | characters/registry.json + 三视图 PNG | Stage 6(三视图调 siliconflow-img-gen) |
slot-plan | storyboard.json + tone | slots/slot-plan.json | Stage 7 |
asset-resolve | slot-plan.json | slots/asset-resolve.json(含 rejected_picks)+ 素材落 raw_materials/ | Stage 8(调 pexels-footage / pixabay-footage / aigc-video-gen) |
slideshow-risk | storyboard.json + slot-plan.json + asset-resolve.json | slots/slideshow-risk.json(六维分) | Stage 9a |
delivery-promise-lock | storyboard.json + brief.md | slots/delivery-promise.json(八类锁) | Stage 9b |
render-shot | shot_decompose.json + characters/ + slot-picks | render/shot-NN/ 下产物 | Stage 10(调 aigc-video-gen i2v / siliconflow-img-gen) |
mix-audio | script.md(delivery_cues) | audio/ 目录 + subtitles.srt 模板 | Stage 11(四场景分流:A 声画同出 / B 旁白一次性 TTS+ASR 对齐 / C BGM 成片后生成(优先 bgm-library 免版税曲库 / pexels/pixabay 并列;定制风格用 aigc-video-gen music)/ D 用户口播录音 → ASR 时间戳 → 按时间戳补素材) |
narration-align | audio/narration.mp3 + audio/narration.subtitle.json | audio/narration-segments.json(统一 segments 格式) | Stage 11b(优先复用 awk-tts --enable-subtitle 落盘的 TTS 原生字级时间戳,缺失时回退火山 ASR 极速版,凭据复用 viral-chaser 同池 VOLC_ASR_*) |
assemble | render/ 顺序 | video.mp4 | Stage 12(按序拼接 + 可选转场 + 可选分辨率归一化 + 自动补静音/统一音频格式) |
add-silent-audio | 无音频视频片段 | 含静音音轨的视频 | Stage 12 原子工具:concat 前补齐音轨,保持连续 |
clip-trim | 素材路径 + 入点/出点/倍速 | 切好的片段 | Stage 12 原子工具:精确切素材,视频和音频分别处理 |
audio-mix | 多条音轨 + 各自延时/音量 | 混合音频 | Stage 12 原子工具:多轨混音 |
timeline-compose | timeline.json(每段素材/入点/出点/倍速/音轨及延时) | 合成片段 | Stage 12 原子工具:按时间轴调 clip-trim + audio-mix 合成 |
motion-audit | video.mp4 + delivery-promise.json | review/motion-audit.json(motion_led 抽查) | Stage 13b(补公共 video-review) |
make-cover | brief.md(标题)+ storyboard 关键帧 | cover.jpg | Stage 14a(调 siliconflow-img-gen) |