Public landing page for a video transcription, AI summary, and mind-map workflow.
This repository is intentionally lightweight. The active service code remains private while deployment credentials, sample media, user content, internal paths, and operational configuration are being removed from publishable materials.
The project line covers a pipeline for turning long-form video into structured review material:
| Stage | Purpose |
|---|---|
| Source ingestion | Accept local video files or video-platform source material after authorization |
| Audio extraction | Prepare audio for speech recognition |
| Transcription | Run Whisper-class speech transcription and segment alignment |
| AI summary | Generate structured summaries, topic sections, and action notes |
| Mind map output | Render summary structure as Markmap-style interactive mind maps |
| Service orchestration | Coordinate API, background jobs, queue/cache, and model calls |
The private implementation currently maps to a FastAPI/Celery/Redis style service with GPU transcription and LLM-assisted summarization.
This public repository is a project pointer, not the active source workspace.
| Repository | Visibility | Role |
|---|---|---|
trans-video |
PUBLIC | Public landing page and sanitized project description |
trans-video-api |
PRIVATE | Current service extraction/API workspace |
02-trans_video |
PRIVATE | Legacy workspace kept as historical reference |
Use this project as an AI tool-product example:
- backend workflow design for long-running media tasks,
- queue-based processing around CPU/GPU-heavy steps,
- transcription-to-summary-to-mind-map product flow,
- prompt/output structuring for readable study material,
- public/private boundary management for user media and deployment data.
It is a support project rather than the main robotics project line.
This repository should not contain:
- API keys, cookies, tokens, service credentials, or model-provider secrets,
- private media, user uploads, customer data, or copyrighted source samples,
- internal hostnames, deployment IPs, queue credentials, or storage paths,
- private prompt logs or production transcripts.
Any future demo should use synthetic or explicitly publishable sample media.
Planned public materials:
- sanitized architecture diagram,
- sample input/output using publishable media,
- short case-study README or NotionNext article,
- screenshot of the generated mind-map output.
Until those are available, this repository should be treated as a clean public pointer to the project line.