KrillinAI

KrillinAI is an open-source audio and video localization stack written in Go. It covers acquisition via yt-dlp or local files, Whisper-family transcription, LLM subtitle segmentation and translation, TTS dubbing, horizontal and vertical composition, and cover images. It ships a desktop client, a local web UI, Docker, and a phased CLI with JSON outputs so scripts and coding agents can call one stage or chain a full pipeline.

Most localization tools stop at subtitles or force a closed SaaS path. KrillinAI is a full open pipeline you can run yourself: pull a video, get accurate captions, LLM translation with context, TTS dubbing with voice options, landscape and portrait renders for the big short-video platforms, plus cover generation. The staged CLI and agent skills matter if you want to automate growth content instead of clicking through a desktop app every time.

Download a release or build the CLI, add API keys for ASR, LLM, and TTS in config, then run desktop, the local web UI, or staged commands like subtitle, tts, render-horizontal, render-vertical, cover, and pipeline. Each CLI stage writes structured artifacts and a manifest so later stages reuse prior work. Agents can also load the per-stage skills under skills/ without scraping CLI docs.

Not a caption-only tool and not a black-box cloud dubber. It is end-to-end localization you own, with platform-aware vertical and horizontal outputs, and a CLI or skill contract built for agent workflows as much as for humans.

• Public demo available

• CrisperWhisper
• OpenMontage
• MOSS Transcribe Diarize