Three Must‑Try Open‑Source Tools: Local Tunneling, AI Short‑Video Creation, and Multi‑Model Switching

tunnelto – Rust‑based local tunnel for exposing services

tunnelto is a lightweight command‑line tool written in Rust that creates a public HTTPS endpoint for a local TCP service without requiring a remote server or DNS configuration. It runs on Windows, macOS and Linux, consumes minimal CPU and memory, and can be started with a single command, e.g. tunnelto 8080 The command opens a tunnel to tunnelto’s cloud service and prints a public URL (e.g., https://abcd.tunnelto.dev) that forwards traffic to the specified local port. The tool also includes an integrated traffic inspector that prints request headers and response bodies in the terminal, eliminating the need for external packet‑capture utilities.

Key technical points:

Implemented in Rust; binary size ~2 MB, memory usage <10 MB.

Supports HTTP/HTTPS forwarding, custom subdomains, and TLS termination.

Cross‑platform binaries are provided for x86_64 and ARM architectures.

Built‑in request/response logger displayed in real time.

Repository: https://github.com/agrinman/tunnelto

Toonflow‑app – End‑to‑end AI pipeline for converting text to short video

Toonflow‑app is an open‑source platform (AGPL‑3.0, version 1.0.6) that automates the full workflow from a raw narrative to a rendered short video. The user supplies a plain‑text story; the system then:

Uses a large‑language model to generate a structured screenplay.

Invokes a text‑to‑image model (e.g., Stable Diffusion) to create character designs and scene assets.

Generates a storyboard by arranging the assets according to the script.

Feeds the storyboard into a text‑to‑video model (e.g., AnimateDiff or similar) to synthesize the final video clip.

The entire pipeline can be run locally, preserving source material privacy. Configuration files allow users to select alternative models, adjust resolution (e.g., 720p), frame rate, and inference steps. Deployment can be performed with a single Docker command or by executing the provided run.sh script.

Technical highlights:

Modular architecture: separate Python packages for script generation, image synthesis, and video rendering.

Supports GPU acceleration (CUDA 11+); CPU fallback is possible but slower.

Provides a simple web UI (FastAPI + Vue) for drag‑and‑drop story input.

All assets are stored in a local output/ directory, enabling reproducibility.

Repository: https://github.com/HBAI-Ltd/Toonflow-app

cc‑switch – Cross‑platform desktop assistant for managing multiple AI model services

cc‑switch is a Tauri 2‑based desktop application (version 3.10.2) that centralizes configuration for several large‑model APIs such as Claude, Codex, OpenCode and Gemini. It runs on Windows, macOS and Linux and provides a graphical interface for:

Storing API keys, endpoint URLs, and model‑specific parameters in a single JSON/YAML file.

Switching the active model with one click, automatically updating environment variables for downstream CLI tools.

Invoking secondary models from within a primary model’s session (e.g., Claude can call Codex for code generation).

Selecting UI language (English, Chinese, Japanese).

Typical workflow:

# Open cc-switch, add a new profile
Profile: Claude
  api_key: ********
  endpoint: https://api.anthropic.com/v1/complete

# Switch to Gemini
Select "Gemini" → active profile updates
# CLI tools now read the active profile automatically

Technical details:

Built with Tauri 2, leveraging Rust for the backend and a webview for the UI.

Configuration stored in ~/.config/cc-switch/config.json by default.

Provides a command‑line helper cc-switch --set <profile> for script integration.

Binary size ~5 MB; supports both x86_64 and ARM64.

Repository: https://github.com/farion1231/cc-switch