Clipping
ai-highlight-clip
Desktop highlight editor with Whisper transcription and LLM clip scoring.
★ 105 starsMIT
View guide →WhisperX — Whisper transcription with word-level timestamps and speaker diarization: 70x realtime batched inference, forced phoneme alignment, BSD-2-Clause
WhisperX is an open-source clipping project by m-bain (m-bain/whisperX on GitHub). It has 24,669 GitHub stars and 2,490 forks, is released under the BSD-2-Clause license, and was last updated on GitHub on 18 October 2026.
The simple route from the README: pip install whisperx (install the CUDA toolkit first if you want GPU acceleration; CPU works without it).
Run whisperx on an audio or video file — it transcribes, aligns word-level timestamps and (with diarization enabled and a Hugging Face token) labels speakers.
Take the JSON/SRT/VTT output into your subtitle, clipping or editing pipeline — word timings and speaker turns are already in the file.
Forced phoneme alignment turns Whisper segments into accurate per-word timings — the feature plain Whisper lacks.
Who said what, labelled per segment, alongside the transcript.
VAD-based batched inference on faster-whisper makes long recordings cheap to process on a GPU.
JSON, SRT, VTT and other subtitle formats that clipping and editing tools consume directly.
The README recommended route.
pip install whisperxThe README fast extra: Numba JIT for the alignment loops on long files.
pip install "whisperx[fast]"Basic CLI use; add --diarize and alignment options per the README for the full pipeline.
whisperx path/to/audio.mp3 --model large-v2Max Bain (m-bain/whisperX on GitHub); the underlying ASR is OpenAI Whisper via faster-whisper, with wav2vec2 alignment models.
Yes — BSD-2-Clause licensed. Diarization models from Hugging Face may ask you to accept their terms and use a token, per the README.
Whisper gives you segment-level text. WhisperX adds accurate word-level timestamps via forced alignment, speaker diarization, and much faster batched inference — the three things production pipelines actually need.
Against faster-whisper alone, WhisperX is the full pipeline (alignment + diarization) rather than just the fast ASR engine — it is built on faster-whisper. Against hosted transcription APIs, it runs locally and free, at the cost of your own GPU time.
pip install whisperx, then whisperx yourfile.mp3 --model large-v2. Add --diarize once you have a Hugging Face token configured.
Desktop highlight editor with Whisper transcription and LLM clip scoring.
FunASR-powered video transcription, subtitles and LLM-assisted clipping with a local Gradio UI.
Free YouTube-to-viral-clip cutter with auto transcription and 9:16 edits.
Auto-download livestreams, AI-clip the highlights, upload in realtime.
Local-first clipper: Whisper plus Ollama turn long video into captioned 9:16 clips.
Traceable AI pipeline for vertical videos across TikTok, Shorts, and Reels.
The most-starred other Clipping projects in the LibreAI catalogue — the realistic shortlist if WhisperX is not the right fit.
Toonflow 是开源 AI 创作平台，融合无限画布、AI Agent 与可视化工作流，支持图像生成、视频生成、智能分镜及短剧创作。支持本地部署、自由接入模型，提供跨平台桌面端，并可通过 MCP 与插件扩展创作能力。Open-source AI creative…
🚀 Truly open-source AI avatar(digital human) toolkit for offline video generation and digital human cloning
macOS video editor built for AI
MoneyPrinter — automate YouTube Shorts from a single topic: local Ollama writes the script and metadata, stock footage and voiceover are assembled into a finished short, MIT
AutoClip｜一个链接，一键出片。开源 AI 视频剪辑桌面工具，将播客、访谈、课程等长视频自动剪成短视频，生成字幕、封面和发布文案，适配抖音、小红书、TikTok、Reels 与 YouTube Shorts。Open-source AI video clipping & content repurposing.
Free and open — grab the source, star the repo, and start creating.