How engrtarun/t_dubber works
Tarun Dubber AI is a Gradio control panel for submitting a video to a Kaggle GPU worker and presenting the downloaded dubbed video and report. The principal flow is upload and project persistence, Kaggle dataset creation/upload, worker dispatch and polling, then output download and display; project history provides a separate browsing capability. The repository also includes a standalone Telegram backup helper, but no shown caller connects it to the dubbing workflow. The worker notebook is documented by pipeline.py as the submitted execution code; its internal processing is not available in the sampled source.
Main entry points: pipeline.py Pipeline coordinator · projects Project files · app.py Gradio control panel · telegram_backup.py Telegram backup helper
Components 17
Studio interface
- Gradio control panel
app.py - Project setup
app.py - Project history
app.py - Report metrics
app.py
Cloud orchestration
- Pipeline coordinator
pipeline.py - Dataset preparation
pipeline.py - Worker dispatch
pipeline.py - Worker polling
pipeline.py - Output retrieval
pipeline.py
Project artifacts
- Project files
projects - Kaggle worker notebook
kaggle_worker.ipynb - Quality report
projects
External services
- Kaggle CLI
- Kaggle GPU
- Telegram backup helper
telegram_backup.py - Telegram API
Other components
- Video uploader
Connections 22
- Video uploader → Gradio control panel: uploads video
app.py - Gradio control panel → Project setup: starts dubbing
app.py - Project setup → Project files: persists video
app.py - Project setup → Pipeline coordinator: calls pipeline
app.py - Pipeline coordinator → Dataset preparation: prepares upload
pipeline.py - Dataset preparation → Project files: writes dataset
pipeline.py - Pipeline coordinator → Kaggle CLI: runs commands
pipeline.py - Dataset preparation → Kaggle CLI: creates dataset
pipeline.py - Pipeline coordinator → Worker dispatch: pushes worker
pipeline.py - Worker dispatch → Kaggle worker notebook: copies notebook
pipeline.py - Worker dispatch → Kaggle CLI: pushes kernel
pipeline.py - Worker polling → Kaggle CLI: checks status
pipeline.py - Pipeline coordinator → Worker polling: waits for completion
pipeline.py - Kaggle GPU → Project files: produces outputs
- Pipeline coordinator → Output retrieval: downloads outputs
pipeline.py - Output retrieval → Kaggle CLI: requests output
pipeline.py - Output retrieval → Project files: saves artifacts
pipeline.py - Gradio control panel → Project history: loads projects
app.py - Project history → Project files: reads files
app.py - Gradio control panel → Report metrics: renders report
app.py - Report metrics → Quality report: reads report
app.py - Telegram backup helper → Telegram API: sends requests
telegram_backup.py