{"slug": "scenemaker-open-models-for-korean-broadcast-highlights-and-what-0-5-fps-costs-in", "title": "SceneMaker: open models for Korean broadcast highlights, and what 0.5 fps costs in video understanding", "summary": "Solbox and SBSi built SceneMaker, an open-model pipeline that turns long broadcast video into short clips, and ran it with Korean broadcaster SBS on Team Korea's baseball games at the 2026 Asian Games, drafting 79 clips that operators reviewed for SBS's \"AI PICK\" page. In testing Qwen3-Omni-30B-A3B-Instruct on 100 six-second Korean news clips, 0.5 fps ran 2.4x faster than 30 fps and cut prompt tokens from 11,791 to 1,711 per clip, though scene changes and continuous action degraded while on-screen text and logos held up. Hand-scoring the final setup at 0.5 fps on 600 clips across 6 genres gave an overall normalized score of 0.93, from 0.98 on news down to 0.91 on baseball, documentary and drama, with on-screen text highest at 0.96 and sounds lowest at 0.89; the full evaluation is still in progress.", "body_md": "Hi everyone, we’re building SceneMaker, which turns long broadcast video into short clips using open models from the Hub. We ran it with SBS, a Korean broadcaster, on Team Korea’s baseball games at the 2026 Asian Games. It drafted 79 clips, operators reviewed them, and they went out on SBS’s “AI PICK” page. The full evaluation is still in progress.\n\nThe main models we use:\n\nEarlier on we also tested how well Qwen3-Omni-30B-A3B-Instruct understands 6-second Korean broadcast clips, and two results might be useful if you work with video.\n\nOn 100 news clips, 0.5 fps was 2.4x faster than 30 fps and prompt tokens dropped from 11,791 to 1,711 per clip. On-screen text and logos held up about the same, but scene changes and continuous action got worse.\n\nWe then scored the final setup by hand at 0.5 fps on 600 clips across 6 genres. Normalized to 0-1, the overall score was 0.93, from 0.98 on news down to 0.91 on baseball, documentary and drama. By field, on-screen text scored highest (0.96) and sounds lowest (0.89).\n\nClips: [SBS 나고야 아시안게임 : AI PICK : SBS](https://programs.sbs.co.kr/sports/ag2026/clips/90399)\n\nDocs, benchmarks and our GitHub: [https://doc.scenemaker.solbox.com](https://doc.scenemaker.solbox.com)\n\nBuilt by Solbox and SBSi under the 2026 NIPA Open-Source AI/SW Development and Utilization Support Program.\n\nHas anyone measured how video understanding changes with frame rate on sports or other action-heavy footage? That’s the part we’d most like to compare notes on.", "url": "https://wpnews.pro/news/scenemaker-open-models-for-korean-broadcast-highlights-and-what-0-5-fps-costs-in", "canonical_source": "https://discuss.huggingface.co/t/scenemaker-open-models-for-korean-broadcast-highlights-and-what-0-5-fps-costs-in-video-understanding/191318#post_1", "published_at": "2026-11-02 01:41:37+00:00", "updated_at": "2026-11-02 01:52:05.772339+00:00", "lang": "en", "topics": ["computer-vision", "ai-research", "machine-learning", "ai-tools"], "entities": ["Solbox", "SBSi", "SBS", "SceneMaker", "Qwen3-Omni-30B-A3B-Instruct", "NIPA", "2026 Asian Games"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/scenemaker-open-models-for-korean-broadcast-highlights-and-what-0-5-fps-costs-in", "markdown": "https://wpnews.pro/news/scenemaker-open-models-for-korean-broadcast-highlights-and-what-0-5-fps-costs-in.md", "text": "https://wpnews.pro/news/scenemaker-open-models-for-korean-broadcast-highlights-and-what-0-5-fps-costs-in.txt", "jsonld": "https://wpnews.pro/news/scenemaker-open-models-for-korean-broadcast-highlights-and-what-0-5-fps-costs-in.jsonld"}}