Projects
dosmoon ships software across creator tools and local AI infrastructure. Each project below has its own GitHub repository; dedicated documentation sites are coming online progressively.
VideoCraft
Section titled “VideoCraft”A program generator where the editor is AI — turn one source video into a finished video or a batch of short clips. A Windows desktop app: ffmpeg ships in the installer, the core features need no account or API key, and AI (ASR / translation / analysis) is an optional three-tier layer (built-in local models / self-hosted aistack / your own cloud keys).
- Repository: github.com/dosmoon/VideoCraft
- Documentation: dosmoon.com/VideoCraft/
- Download: latest Windows installer
- Capabilities: Clip (batch short videos from subtitle hotclips) · News Desk Video (bilingual subs + name plates + chapter strip) · full subtitle pipeline (ASR / import / translate / QC auto-fix / styled burn-in)
- Stack: Electron, Python
- Platform: Windows 10 / 11, 64-bit
aistack
Section titled “aistack”Local-first ASR (speech recognition) and TTS (speech synthesis) service exposing an OpenAI-compatible HTTP API. Built to be the speech-side companion to Ollama on the LLM side — point any tool that already speaks the OpenAI API at localhost and get local speech.
- Repository: github.com/dosmoon/aistack
- Documentation: dosmoon.com/aistack/
- Capabilities: faster-whisper / Parakeet / SenseVoice ASR backends · Qwen3-TTS via vLLM-Omni · per-capability rolling p50/p95/p99 metrics · cross-backend WER/RTF benchmarking
- Stack: Python, vLLM, Docker
- Positioning: Research-grade — built for developers who want to compare backends and integrate local speech into their own pipelines, not for end users expecting a one-click installer.
More projects will land here as they go public.