動畫 / VFX / 虛擬製作

讓中小團隊隨時看到演出效果

思看(Semcam Live) 面向動畫、遊戲、VFX、虛擬拍攝和數位角色製作，把邊緣 AI 無標記人體捕捉、Goku 光學剛體追蹤、HPE 高精度處理和下游軟體接口放進同一條輕量製作鏈路。

核心價值

行業裡的無標記動捕和即時製作系統說明了一個很明確的趨勢：創意團隊希望在早期構思、預演和拍攝現場就能看到角色動作，而不是等演員穿好動捕服、完成貼點、排隊進入動捕棚之後才知道表演是否成立。大型棚拍依然有價值，但對中小製作團隊、遊戲團隊、短劇和虛擬製作團隊來說，動捕棚檔期、人員協調、技術維護和預算都會限制迭代速度。思看(Semcam Live) 的定位是把動作捕捉從「預約一次昂貴拍攝」變成「隨時可以打開的創意工具」：演員無需動捕服或貼點即可表演，導演和動畫師現場看到角色效果；道具、武器、相機、手柄和其他剛體物件由 Goku 光學系統同步追蹤；關鍵鏡頭再透過 HPE 高精度處理獲得更乾淨的資料。

適用場景

01

角色動畫與動作預演

Animators, directors, and action directors can capture performer motion during rehearsal, review character rhythm, center of mass, blocking, and emotional expression on site, and make creative decisions earlier.

02

遊戲動作與招式設計

For combat moves, skills, NPC behavior, cutscenes, and multiplayer interaction prototypes. Teams can try, record, and revise motion frequently without entering a full stage-shoot workflow every time.

03

虛擬拍攝與即時預覽

Send performer motion into UE/Unity in real time with virtual sets, virtual cameras, and character assets so directors can judge shots, composition, pacing, and performance relationships on site.

04

道具、武器與攝影設備

Goku synchronously tracks swords, firearms, canes, rackets, controllers, cameras, stabilizers, and other rigid objects so human performance and prop state share one coordinate system.

05

中小團隊自建動捕空間

Fixed PoE cameras, simplified calibration, and a local data chain make it easier for small studios, game teams, schools, and content teams to own their own motion capture capability.

06

關鍵鏡頭採後精修

Real-time preview supports fast decisions, while HPE high-precision processing refines key shots and complex actions after capture, letting lightweight capture and high-quality delivery coexist in one workflow.

AI 無標記與光學剛體的融合工作流

01

演員無需動捕服和貼點

思看(Semcam Live) captures human motion with fixed edge AI cameras, reducing performer preparation time so actors can enter character more naturally and creative teams can start auditions and previs faster.

02

光學系統同步追蹤道具

Goku tracks props, weapons, camera gear, controllers, interactive devices, and other rigid objects, solving the difficulty pure markerless human capture has in expressing stable object poses.

03

現場即時看到角色效果

120fps motion output and latency below 0.1 seconds can feed real-time engines, letting directors, animators, and clients see driven character results on site and reducing rework after capture.

04

下游軟體接口完整

Supports Unreal Engine, Unity, Live Link, FBX, BVH, Blender, Maya, MotionBuilder, C++, Python, and related workflows, making it easier to connect existing rigs, retargeting, and animation cleanup pipelines.

05

標定採集更輕

Fixed camera arrays and local Active Center management lower setup friction, helping teams complete camera management, joint calibration, performer capture, prop tracking, and data export faster.

06

技術維護更便捷

Edge-side processing handles complex image features while the central system receives structured data and reconstructs motion, reducing massive video-stream transmission pressure and making a long-term mocap space easier for small teams to maintain.

與傳統動捕棚的關係

Traditional optical mocap stages remain important infrastructure for high-end film, games, and virtual production, especially when a project tightly controls space, performers, props, and shooting rhythm. But they often mean booking stage time, coordinating performers, preparing suits and markers, waiting for technical setup, and then handing data to postproduction. 思看(Semcam Live) is better suited as a team-owned permanent mocap capability: teams can try actions, block shots, validate characters, and build motion libraries every day, then move into heavier production only when premium stage capture or final data refinement is needed.

為什麼要 AI + 光學融合

Pure AI markerless capture is good for fast human motion, but props, weapons, cameras, controllers, and interactive objects need more stable rigid-body poses. Pure optical capture is good for high-precision object tracking, but if every performer relies on suits, markers, or marker sets, preparation and maintenance costs rise. 思看(Semcam Live) + Goku divide the work: bodies are captured by AI markerless cameras, key objects are tracked as optical rigid bodies, and both share one coordinate system and timeline.

從預演到交付的資料鏈

In early creative stages, real-time markerless capture helps teams see character motion quickly. During rehearsal and shooting, Goku synchronously records props and device poses. For key shots, HPE high-precision processing improves motion data quality. At delivery, FBX, BVH, Live Link, UE, Unity, Blender, Maya, MotionBuilder, C++, Python, and related interfaces carry data into downstream production.

建議使用產品型號

根據您的場景規模和需求，我們推薦以下配置

Semcam Live

PRO

面向 4 人以內即時無標記捕捉的邊緣 AI 動捕相機，內建 40 TOPS 嵌入式 GPU，支援 120fps 即時輸出。

適合小型工作室、教學展示、內容預演與快速部署

內建算力

40 TOPS

系統功耗

10W

最大解析度

230萬像素

最大捕捉人數

4人

最遠追蹤距離

18m

關節3D旋轉角

支援

手指捕捉

—

臉部捕捉

—

HPE

—

Semcam Live

PRO+

面向高品質動作資料採集的專業邊緣 AI 動捕相機，內建 55 TOPS 嵌入式 GPU，支援 HPE 高精度解算擴充。

適合生命科學、高校實驗室、機器人訓練與動畫/VFX

內建算力

55 TOPS

系統功耗

15W

最大解析度

230萬像素

最大捕捉人數

4人

最遠追蹤距離

18m

關節3D旋轉角

支援

手指捕捉

✓ 支援

臉部捕捉

—

HPE

✓ 支援

客製化方案

不確定哪款適合您？

聯絡我們的技術團隊，根據您的場景規模、人數、空間與精度要求，客製化最佳相機陣列方案。

把動捕從「預約棚期」變成「每天都能試」。

Contact the 思看(Semcam Live) team to plan an AI + optical fusion motion-capture workflow for animation, games, VFX, and virtual production based on your team size, capture space, character assets, prop types, engine plan, and postproduction software pipeline.

FAQ

常見問題

產品行業應用

思看(Semcam Live) 是什麼？

思看(Semcam Live) 是邊緣運算 AI 無標記動作捕捉系統。相機側完成目標偵測、2D 關鍵點、ReID、追蹤、分割、裁切和置信度計算，Active Center 在本地完成跨相機匹配、三角化、3D 關鍵點、骨架解算、IK、濾波、retarget 和導出。

行業應用工作流程

人體捕捉需要穿戴裝置或貼點嗎？

人體動作捕捉不需要動捕服、慣性感應器或反光標記點。若專案需要追蹤剛體對象，例如工具、道具、機器人末端、手柄或訓練器具，可以使用 Goku 進行 6DoF 剛體追蹤。

產品行業應用工作流程

即時鏈路和 HPE 高精度解算拓展分別適合什麼？

即時連結提供 120fps 動作輸出和低於 100ms 的端對端延遲，適合現場預覽、即時驅動、互動和擷取檢查。 HPE 是採後高精度處理流程，適合研究、訓練資料生產、關鍵鏡頭和需要更高品質關鍵點結果的專案。