ShortForge

AI video-to-shorts desktop app — converts 1–10 GB videos into scene-aware short clips with multimodal LLM scene understanding and auto captions.

ShortForge is a Windows desktop application for turning long videos into scene-aware short clips. It supports an analyze-first workflow: import a video, detect scene changes, review the expected segments, choose an output ratio and reframing mode, then process and export selected clips. Users can rename clips, preview them and cancel processing with stage-level progress. The Electron and Node.js application orchestrates FFmpeg and FFprobe. Large source files are read from disk rather than loaded in full into memory. Segmentation works toward a target duration and snaps cuts to scene boundaries within a minimum–maximum range. When no suitable boundary is available it splits at the maximum; a short final tail is merged with the previous segment. This balances scene continuity with useful clip lengths while keeping the segmentation intended to cover the source timeline. Encoding detects NVIDIA NVENC, Intel QSV or AMD AMF and falls back to CPU processing. Output choices include blurred padding and center cropping. Optional local Whisper transcription supports burned-in captions and transcript-aware titles; if that dependency is unavailable, the core cutting workflow remains usable. The optional OpenRouter AI layer adds scene descriptions, ranking scores, title/caption/hashtag suggestions and smart-crop hints. Core video processing is local, but enabling external AI sends selected inputs to a provider and requires network access and an API key. The key is stored using Electron safeStorage with Windows DPAPI and used from the main process. Generated rankings are suggestions for clip selection, not evidence of future audience reach.

Electron, Node.js, FFmpeg, OpenRouter, Whisper

Desktop · AI Video