103 lines
2.5 KiB
Plaintext
103 lines
2.5 KiB
Plaintext
|
# TEST-LINE-001-SUBAI · RUN-001报告
|
|||
|
|
|
|||
|
|
> HLDP://video-ai-system/outputs/tests/TEST-LINE-001-SUBAI-RUN-001-REPORT
|
|||
|
|
> 类型: 角色台词/口型突破测试 · 火山Seedance
|
|||
|
|
> 生成: D143c · 2026-06-23 21:54 CST
|
|||
|
|
> 铸渊 ICE-GL-ZY001
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 原文锁定
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
苏白（大声）：未来的天下第一宗！天道宗开门收徒啦！
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
不得改写，不得替换为旁白。
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 输入
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
model: doubao-seedance-2-0-260128
|
|||
|
|
task_id: cgt-20260623215402-9zn77
|
|||
|
|
duration_request: 4s
|
|||
|
|
resolution_request: 720p
|
|||
|
|
reference_image: /Volumes/JZAO/铸渊-ICE-GL-ZY001/OUT-输出/图片/zai-fu-fei-xiu-xian/ep01/anchors/char-003-subai-3d-reference-v2.png
|
|||
|
|
output: /Volumes/JZAO/铸渊-ICE-GL-ZY001/OUT-输出/视频/zai-fu-fei-xiu-xian/ep01/tests/TEST-LINE-001-SUBAI-RUN-001.mp4
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
提示词目标:
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
苏白本人中近景，双手叉腰，大声喊原文台词。
|
|||
|
|
必须像角色本人说话，不是旁白。
|
|||
|
|
不生成字幕。
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 客观探针
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
container_duration: 4.086009s
|
|||
|
|
video_codec: h264
|
|||
|
|
video_size: 834x1112
|
|||
|
|
fps: 24
|
|||
|
|
audio_codec: aac
|
|||
|
|
audio_rate: 44100Hz stereo
|
|||
|
|
audio_mean_volume: -19.3 dB
|
|||
|
|
audio_max_volume: -4.1 dB
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 画面观察
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
pass:
|
|||
|
|
- 竖屏成立。
|
|||
|
|
- 画质明显优于preview-002。
|
|||
|
|
- 苏白白衣少年形象基本保持。
|
|||
|
|
- 双手叉腰姿势成立。
|
|||
|
|
- 嘴部有明显开合，具备“角色在说话”的画面基础。
|
|||
|
|
- 背后天道宗牌匾未完全漂移，镜头内文字可识别。
|
|||
|
|
|
|||
|
|
risk:
|
|||
|
|
- 发型/脸部仍有轻微模型再创作，不能直接批准为跨集CHAR最终资产。
|
|||
|
|
- 生成视频自带音轨，但尚未验证其是否按原文台词准确发声。
|
|||
|
|
- 未使用独立角色对白音频驱动口型，因此不能判定“配音/口型问题已解决”。
|
|||
|
|
- 背景牌匾只是随镜生成参考，不能代替PROP-TDZ-PLAQUE正式贴图流程。
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 裁决
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
status: PARTIAL_PASS
|
|||
|
|
approved_for: TEST-LINE-001后续视觉口型对照
|
|||
|
|
not_approved_for: 成片、最终配音、最终口型、最终CHAR资产
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
结论:
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
火山Seedance可作为“苏白开口表演画面”的主模型候选继续测试。
|
|||
|
|
角色对白必须另建Voice/LipSync链路，不能继续用旁白或不可验证音轨替代。
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 下一步
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
1. 生成/确认苏白最终正面、侧脸、全身、张嘴表情资产。
|
|||
|
|
2. 建立VOICE-CHAR-003的真实角色对白适配器。
|
|||
|
|
3. 用同一句原文生成独立对白音频。
|
|||
|
|
4. 用对白音频驱动画面口型，或对Seedance成片进行后置口型同步。
|
|||
|
|
5. 以原文逐字准确、嘴型同步、情绪像苏白本人为通过标准。
|
|||
|
|
```
|
|||
|
|
|