ai-media
bowen31337 제작
DevOps & Cloud요약
Full-stack AI media generation powered by GPU server (RTX 3090/3080/2070S).
설명
이 스킬이 하는 일
Runs image, video, talking head, and voice synthesis jobs on a remote GPU server over SSH. Scripts wrap ComfyUI, SadTalker, and Voxtral pipelines and return output file paths when generation completes. Supports multiple models per media type with configurable style, duration, and language parameters.
Offloads GPU-intensive generation to a dedicated server so the local machine stays free and jobs run in the background without blocking other work.
사용 시점
- Generating photorealistic images from text prompts for mockups or content
- Creating short AI-generated videos for social media or product demos
- Animating a static portrait photo into a lip-synced talking head video
- Synthesizing voiceover audio in multiple languages and voice genders
- Producing agent-narrated demo videos with an animated avatar and synced speech
예시 워크플로
User asks: Make a talking head video of someone saying 'Welcome to the future' in a neutral voice using photo.png
- 1
Verify SSH connectivity to the GPU server using the configured key
- 2
Call talking-head.sh with the speech text, 'neutral' voice style, and photo.png as the avatar
- 3
SadTalker processes audio-driven animation on the GPU server
- 4
Script waits for completion and retrieves the output file path
Path to the generated video, e.g. /data/ai-stack/output/talking_001.mp4, ready in roughly 30-40 seconds for a 10-second clip
요구 사항
- GPU server accessible via SSH with GPU_USER and GPU_HOST environment variables configured
- SSH private key present at ~/.ssh/id_ed25519_gpu with access to the GPU server
설치
clawhub install bowen31337/ai-mediaCLI 명령을 실행하거나, 프롬프트를 Claude Code, Cursor, Codex 등 사용하는 코딩 에이전트에 붙여넣으세요.