TensorPlayer - Official Desktop Application
Breaking the single-window paradigm: A multimodal media workstation powered by fluid physics layout and tensor-grade computer vision.
打破传统播放器的“单一视窗”牢笼 —— 基于流体物理排版与计算级 AI 的多模态多媒体工作台。
For decades, mainstream media players have adhered to a monolithic paradigm: a black frame, a static seek bar, and the linear playback of a single media file. However, modern workflows demand far more:
- Multimodal Comparison: Synchronizing multi-angle camera feeds alongside text documentation, code, and reference imagery.
- Micro-Analysis & Review: Looping multiple non-contiguous segments inside a single lecture, dance rehearsal, or surveillance feed.
- In-Stream Vision Inspection: Extracting raw pixel channels, enhancing visibility, or tracking moving targets in real-time without re-rendering in external NLEs.
iClovers Tensor Player breaks the "single-window cage." It is not merely a player, but a multimodal vision workbench combining a fluid layout engine, multi-range loop control, and tensor-level computer vision pipelines.
- 🧩 Fluid Physical Layout Engine
- Responsive Grid: Built on an automated pathfinding and physical-avoidance layout algorithm with automatic compaction.
- Freeform Canvas: Pack selected media windows into floating, zoomable canvas boards for flexible multi-window arrangements.
- 🔁 Multi-Range A-B Loop Playback
- Eliminates the constraint of conventional single-pair
$[A, B]$ repeat loops. - Mark multiple distinct time segments across the timeline; the playback engine automatically skips intermediate gaps, stitches segments seamlessly, and loops continuously.
- Eliminates the constraint of conventional single-pair
- 📺 Surveillance-Grade Video Wall Matrix
- Instant switching across 2x2, 3x3, center-focused, and absolute seamless stitching layouts.
- Sub-Grid Nesting: Automatically folds overflowing channels into nested sub-grids without breaking geometric layout constraints.
- Native support for RTSP streaming, ONVIF PTZ cameras, and automated directory radar monitoring.
- 👁️ Real-Time Tensor-Grade Vision Algorithms
- Image Enhancement: Dark Channel Prior (DCP) dehazing, adaptive low-light enhancement, Retinex SSR, and CLAHE.
- Color Channel Decomposition: Split running video streams into synchronous 2x2 / 3x2 matrices of RGB, YUV, HSV, CIELAB, or color-blindness simulation modes.
- Live Target Perception: CSRT tracking (with auto-framing and zoom-in tracking), YOLO object detection, OCR subtitle extraction, and instant QR/barcode decoding.
- 🛠️ End-to-End Post-Production Loop
- Attach auxiliary background audio tracks, segment-based progress bar watermarks (.pgwm), and vector ASS subtitles.
- Powered by embedded FFmpeg pipelines for transition-rich rendering (xfade/acrossfade), format remuxing, and frame extraction directly within the workstation.
几十年来，主流播放器始终停留在“一个黑框、一条进度条、按部就班播放一个文件”的传统范式。但在实际的创作与技术工作中：
- 素材对比：需要一边核对机位画面，一边查阅工程文档和参考图；
- 片段精读：想要针对同一个长视频圈选多个不同时间段进行跳跃、反复循环复盘；
- 视觉分析：面对恶劣光线或特定通道信息，传统播放器无法在播放中实时拆解与增强像素底层信息。
iClovers TensorPlayer 颠覆了这一现状——它不仅是一个高性能的跨格式播放器，更是一个融合了流体网格排版、多区间循环引擎与**工业级视觉计算（Tensor）**的多模态视讯工作台。
- 🧩 动态流体物理排版引擎
- 自研弹性网格：媒体拖拽自动寻路、物理避让与紧凑紧缩（Compact 算法）。
- 自由画布（Canvas）：支持将任意数量的音视频、文本及网页组合一键升格为悬浮画布，实现无缝拖拽与图层管理。
- 🔁 多区间独立循环（Multi-Range A-B Loop）
- 突破传统播放器单次只能设一组
$[A, B]$ 的限制。 - 支持在单条时间轴上圈选多个时间片段，自动跳跃、平滑衔接并整体循环，专为动作拆解、舞蹈扒舞、网课精听设计。
- 突破传统播放器单次只能设一组
- 📺 专业级安防电视墙矩阵
- 支持 2x2、3x3、中心主讲与绝对无缝拼接等多种矩阵布局。
- 防溢出嵌套算法：超出显示限制时自动将多余画面折叠为子电视墙，严防溢出崩溃。
- 支持 RTSP 网络串流、ONVIF 摄像机接入与文件夹动态雷达监视。
- 👁️ 实时 Tensor 视觉与算法解构
- 画质增强：暗通道先验去雾（DCP）、自适应夜视、Retinex 反射率增强与自适应直方图（CLAHE）。
- 多维色彩解构：播放中实时拆解为 RGB、YUV、HSV、CIELAB 及色盲模拟矩阵。
- 智能视讯感知：内置 CSRT 目标跟踪（支持自动平滑跟拍放大）、YOLO 目标检测、实时 OCR 字幕抓取与二维码秒级识别。
- 🛠️ 从“观看”到“产出”的后期闭环
- 支持注入外挂第二背景音频轨、分段进度条水印（.pgwm）与 ASS 矢量字幕渲染。
- 内置基于 FFmpeg 管道的视频合并、转场特效渲染（xfade/acrossfade）与高帧率无损图片序列切解。