https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's tech blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, distributed training, and data engineering. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Sat, 16 May 2026 00:00:00 GMT https://www.youngju.dev/blog/culture/2026-05-16-ai-video-generation-2026-sora-2-veo-3-runway-gen-4-pika-luma-dream-machine-kling-hailuo-hunyuan-deep-dive.en https://www.youngju.dev/blog/culture/2026-05-16-ai-video-generation-2026-sora-2-veo-3-runway-gen-4-pika-luma-dream-machine-kling-hailuo-hunyuan-deep-dive.en Two years after the 2024 Sora preview stunned the world, AI video generation in 2026 is a different landscape. OpenAI Sora 2 arrived with 4K, 25-second clips, native synchronized audio, and Cameos — alongside an iOS app. Google Veo 3 took Veo 2 a step further with native dialogue sync and sound effects. Runway Gen-4 entered film production with multi-shot consistency, and Pika 2.5 redefined clip editing via Pikadditions · Pikaffects · Pikaframes. Meanwhile open source — Tencent Hunyuan Video (13B Apache-style), Alibaba Wan 2.1, Genmo Mochi 1 (Apache 2.0 10B), Lightricks LTX-Video (real-time 2B) — made cinematic generation possible on local H100/H200/RTX 4090. This guide covers 14 closed and open models, pricing, workflows (t2v · i2v · v2v · lip sync), Korean/Japanese markets, and C2PA watermarking in a single arc. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) ai-video-generationsoraveorunwaypikalumaklinghailuohunyuan-videomochiwanenglish https://www.youngju.dev/blog/culture/2026-05-16-ai-video-generation-2026-sora-2-veo-3-runway-gen-4-pika-luma-dream-machine-kling-hailuo-hunyuan-deep-dive.ja https://www.youngju.dev/blog/culture/2026-05-16-ai-video-generation-2026-sora-2-veo-3-runway-gen-4-pika-luma-dream-machine-kling-hailuo-hunyuan-deep-dive.ja 2024年のSoraプレビューが世界を驚かせてから2年、2026年のAI動画生成はまったく違う風景になっている。OpenAI Sora 2が4K · 25秒 · 同期オーディオ · CameosとiOSアプリを引っ提げて登場し、Google Veo 3はVeo 2からさらに一歩 — ネイティブのダイアログ同期と効果音を一緒に生成する。Runway Gen-4はマルチショット一貫性で映画制作ワークフローに踏み込み、Pika 2.5はPikadditions · Pikaffects · Pikaframesでクリップ編集を再定義した。一方オープンソース陣営ではTencent Hunyuan Video(13B Apache風)、Alibaba Wan 2.1、Genmo Mochi 1(Apache 2.0 10B)、Lightricks LTX-Video(リアルタイム2B)が登場し、ローカルのH100/H200/RTX 4090でもcinematicな動画が可能になった。本稿は14個のクローズド/オープンモデル · 価格 · ワークフロー(text2video · image2video · video2video · リップシンク) · 日本と韓国の市場 · C2PAウォーターマークまでを一気に整理する。 Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) ai-video-generationsoraveorunwaypikalumaklinghailuohunyuan-videomochiwan日本語 https://www.youngju.dev/blog/culture/2026-05-16-ai-video-generation-2026-sora-2-veo-3-runway-gen-4-pika-luma-dream-machine-kling-hailuo-hunyuan-deep-dive https://www.youngju.dev/blog/culture/2026-05-16-ai-video-generation-2026-sora-2-veo-3-runway-gen-4-pika-luma-dream-machine-kling-hailuo-hunyuan-deep-dive 2024년 Sora 프리뷰가 세상을 놀라게 한 지 2년, 2026년의 AI 비디오 생성은 완전히 다른 풍경이다. OpenAI Sora 2가 4K · 25초 · 동기 오디오 · Cameos를 들고 iOS 앱과 함께 등장했고, Google Veo 3은 Veo 2에서 한 단계 더 — 네이티브 다이얼로그 동기화 · 음향 효과를 함께 생성한다. Runway Gen-4는 멀티샷 컨시스턴시로 영화 제작 워크플로에 진입했고, Pika 2.5는 Pikadditions · Pikaffects · Pikaframes로 클립 편집을 재정의했다. 한편 오픈소스 진영에서는 Tencent Hunyuan Video(13B Apache-스타일), Alibaba Wan 2.1, Genmo Mochi 1(Apache 2.0 10B), Lightricks LTX-Video(실시간 2B)가 등장하면서 로컬 H100/H200/RTX 4090에서도 cinematic 영상이 가능해졌다. 이 글은 14개의 클로즈드/오픈 모델 · 가격 · 워크플로(text2video · image2video · video2video · 립싱크) · 한국·일본 시장 · C2PA 워터마크까지 한 호흡으로 정리한다. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) ai-video-generationsoraveorunwaypikalumaklinghailuohunyuan-videomochiwan