tt/ttstd-infinite-canvas
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor(*): 稳定期治理 - 收敛核心架构、错误语义并修复跨项目数据竞态

Browse Source
This commit is contained in:
ddCat
2026-09-09 22:02:10 +08:00
parent bacd502f5c
commit b5f3413ded
298 changed files with 12681 additions and 26428 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
backend/internal/protocol/docs/newapi-channel-1.md
View File

@@ -21,7 +21,7 @@ Content-Type: application/json
{{PARAMETERS}}
实际映射：图片 -> `{type:"reference_image"}`，视频 -> `{type:"reference_video"}`，音频 -> `{type:"reference_voice"}`；均放入 `input.media`。默认 `resolution=720P`、`ratio=16:9`、`duration=5`。只有 `watermark=true` 时才发送水印字段；`generateAudio` 当前不发送。
实际映射：`input.media[].type` 取素材 `role`，缺省时图片为 `reference_image`、视频为 `reference_video`、音频为 `reference_voice`。宿主视频任务会给音频带上 `reference_audio`，因此生产请求里音频 type 是 `reference_audio`。默认 `resolution=720P`、`ratio=16:9`、`duration=5`。只有 `watermark=true` 时才发送水印字段；`generateAudio` 当前不发送。
## 多模态创建示例