lovstudio/dsh-video-studio0

@lovstudio/dsh-video-studio

A local-first video editing workbench for DeepSeek Harness with Remotion, GSAP and pluggable ASR

AI 분석

核心用途是在 DSH 内部进行视频剪辑与制作。适合需要结合 AI 能力、利用代码和时间线工具进行本地视频编辑与渲染的用户。

패키지
@lovstudio/dsh-video-studio
버전
0.1.0
라이선스
MIT
최근 업데이트
2026. 9. 5.

설치

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:lovstudio/dsh-video-studio

Service API, configuration, and extension points

All HTTP routes live under /video-studio/. The DSH host checks the connection service's requestRejection() before serving either API or static files. The browser has no API secrets. Media reads support byte ranges and HEAD for seeking.

EndpointBehavior
GET api/capabilitiesActual ASR and rendering availability
GET api/projectsSaved projects
PUT api/projects/:idValidate and atomically save a project
POST api/assetsStream a media upload into the managed asset store
GET media/:idAuthenticated managed media, including Range requests
POST api/renderQueue an immutable project snapshot for MP4 export
POST api/asrQueue a transcription for an uploaded asset
GET api/jobs/:idJob state and progress
POST api/jobs/:id/cancelCancel queued or running work
GET exports/:id.mp4Download a completed render

For asset upload, the body is the binary file; query fields describe name, kind, duration, and optional width/height. Projects use versioned frame-based data. The host accepts only managed asset references for rendering. Each render receives a temporary loopback media server exposing its snapshot's asset allowlist, without giving Chromium the user's DSH login cookie.

ASR requires explicit server configuration. Choose a provider/model supporting timestamped segments:

export DSH_VIDEO_ASR_ENDPOINT='https://your-provider.example/v1/audio/transcriptions'
export DSH_VIDEO_ASR_MODEL='your-timestamp-capable-model'
export DSH_VIDEO_ASR_FORMAT='diarized_json' # or verbose_json