https://dailypypy.org/tags/timeline/ Recent content in Timeline on 每日拍拍 Hugo -- gohugo.io zh-tw © 2026 每日拍拍 Tue, 22 Sep 2026 11:52:00 +0800 155076163427069952 154825760438254592 https://dailypypy.org/learn/mlx-vlm-video-frame-analysis/ Tue, 22 Sep 2026 11:52:00 +0800 https://dailypypy.org/learn/mlx-vlm-video-frame-analysis/ <!--- 1440x768 prompt: masterpiece, best quality, highres, clean anime illustration, japanese anime style, soft shading, flat color design, 1girl, black hair, green eyes, white off-shoulder shirt, black short skirt, from above, full body, walking lightly along a curved trail of floating blank translucent film frames, one hand making a peace sign, playful wink, looking at viewer, pastel aqua and blush pink background, subtle evenly spaced glowing dots suggesting a timeline without symbols, logos, markings, or text, neat composition, detailed eyes, cute and smart vibe, minimal background, polished illustration, no text negative prompt: worst quality, bad eye, bad hand, extra limbs, manga, multiple views, monochrome, text, signature dedup note: Planned 2026-09-22 topic `mlx-vlm-video-frame-analysis` passed a strict 169-post inventory and current-source scan. ../mlx-vlm-image-chat/ covers interactive single/multi-image questions, preprocessing, chat templates, and vision cache; ../mlx-vlm-batch-structured-output/ covers independent image manifests, constrained JSON, resumability, and batch QA; ../mlx-lm-batch-inference/ is text-only. This follow-up instead extracts timestamped video frames with FFmpeg, creates per-frame evidence, merges adjacent observations into a timeline, builds an event index, and validates temporal coverage and ordering. It does not repeat the older installation, chat, or generic image-batch sections. source check: Reviewed against the official MLX-VLM README and current package metadata plus the FFmpeg filter and ffprobe documentation on 2026-09-22. PyPI lists mlx-vlm 0.7.2. Native video support remains model-dependent, so this tutorial deliberately uses explicit timestamped image frames and the OpenAI-compatible multimodal server contract. ---> <p> <figure> <img class="my-0 rounded-md" loading="lazy" alt="featured" src="./featured.png" /> </figure> </p>