查人Find people

证据：Evidence: 项目公开 3,483 条资产：服装 1,753 条、场景 434 条，造型与舞台动作是这部 2 分钟短片的主体。3,483 public project assets: 1,753 costume (dress/suit) and 434 locations. Styling and stage action carry this 2-minute film.

证据：Evidence: 两部作品场景资产分别 1,679 与 1,574 条、角色资产 1,107 与 488 条——场景一项两部都成立，标「强」。Scene assets of 1,679 and 1,574 across two works, with 1,107 and 488 character assets. Set design holds across both, hence the Strong mark.

证据：Evidence: 在册最长作品（22 分 01 秒）；角色资产 269 条、环境资产 114 条，群像与世界规模成体系。The longest work on record (22:01). 269 character and 114 environment assets: an ensemble and a world, built as a system.

证据：Evidence: 项目资产总量 4,754 条，在册最大体量之一；主用角色向模型 Soul V2，撑起 10 分半叙事片。4,754 project assets, among the largest on record. Built mainly on the character-focused Soul V2, sustaining a 10.5-minute narrative film.

证据：Evidence: 资产按编号分夹管理：分镜 186 条单独成夹，服装 10 条、角色 51 条各有专夹。Assets managed in numbered folders: 186 storyboard assets in their own folder, with dedicated folders for outfits (10) and characters (51).

证据：Evidence: 资产 1,409 条按角色／场景／道具／素材／音乐／音效六夹管理，其中音效 120 条单独成夹。《CONTROL》15 场戏机位与焦点全程锁定。1,409 assets in six managed folders (character, location, props, footage, music, SFX); the 120 SFX assets sit in a folder of their own. On 'CONTROL', camera angles and focus stay locked across all 15 scenes.

证据：Evidence: 制作栈公开可查：使用 Higgsfield Genjutsu 动作迁移与物体替换完成这部 4 分钟动画短片。该片获 Higgsfield 全球电影节观众选择奖（Dae Hyeong Kang 署名，社区投票，10 万美元）。A public, checkable stack: Higgsfield Genjutsu motion transfer and object swap drive this 4-minute animated short. The film won the Audience Choice award at the Higgsfield Global Film Festival (credited to Dae Hyeong Kang, decided by community vote, $100,000).

证据：Evidence: 场景资产 1,462 条、角色 327 条；海报资产 196 条单独成规模，主视觉产出有据。1,462 scene assets and 327 character assets; 196 poster assets form a body of key-art work in their own right.

证据：Evidence: 音频资产 130 条、海报 291 条、演员 11 条各成专夹，声音与主视觉投入均有据。Dedicated folders for audio (130), posters (291) and cast (11): the investment in sound and key art is on the record.

证据：Evidence: 项目将 298 条生成视频按五幕架构分夹管理，并配置 48 个场景（含 26 个家庭室内）与 40 条配乐资产。298 generated video clips organized across a five-act folder structure, supported by 48 scene assets (including 26 home interiors) and 40 score assets.

证据：Evidence: 项目公开的 1,048 项生成资产中，专门配置了 200 项音效（SFX）资产。Among the project's 1,048 public generated assets, 200 are specifically allocated to sound effects (SFX).

证据：Evidence: 角色设定表按身体部位、招牌手势、服装、色彩和姿势分项并附双重锁定清单，同时将创伤、抑郁、悲伤与愤怒四个心理领域实体化为独立生物与环境规则。The character design sheets categorize body parts, signature gestures, outfits, colors, and poses with dual lock lists, while four psychological realms (trauma, depression, grief, and anger) are materialized into distinct creatures and environmental rules.

证据：Evidence: 项目为主角构建了七个生命阶段版本与 409 项角色资产，配置附带空间布局锁定图的 572 个房屋场景资产，并包含 649 个场景分镜资产且全片无对白无配乐。Seven life-stage versions and 409 character assets for the protagonist, 572 house scene assets with spatial layout lock diagrams, and 649 scene storyboard assets, with zero dialogue or score throughout.

证据：Evidence: 项目管理着 340 个场景资产与 61 项声音资产，并将 466 条视频按编号场次分夹，另设 158 个片头与 31 个片尾专夹。340 scene assets and 61 sound assets, organizing 466 videos into numbered scene folders alongside dedicated folders for intro (158) and outro (31).

证据：Evidence: 在 2,196 项项目资产中，包含 930 个场景资产、341 个角色资产以及 107 项主视觉 Key Art。Across 2,196 project assets, the work contains 930 scene assets, 341 character assets, and 107 Key Art visual assets.

证据：Evidence: 项目资产包含 117 个场景，核心制作工具采用场景向的 Cinematic Studio Soul Location 与 Seedance 2.5。The project assets include 117 scenes, utilizing the location-focused Cinematic Studio Soul Location alongside Seedance 2.5 as primary tools.

证据：Evidence: 制作管线整合了 25 种工具，包含 Cinematic Video Editor、Cinematic Video Relight（重打光）、Motion Transfer 以及 Topaz 超分等后期合成流程。The production pipeline integrates 25 tools, featuring post-production and compositing workflows including Cinematic Video Editor, Cinematic Video Relight, Motion Transfer, and Topaz upscaling.

证据：Evidence: 作品时长约 10 分钟，作为完整短片入选参与 Higgsfield Global Film Festival 2026。The work runs approximately 10 minutes, entered as a complete short film in the Higgsfield Global Film Festival 2026.

证据：Evidence: 团队自研 AFX 流程完成 3 分钟短片《One More Pumpkin》，并获 2024 首届迪拜 AI 电影节 Grand Prize 与 Audience Award。The team utilized their proprietary AFX pipeline to complete the 3-minute short 'One More Pumpkin', winning the Grand Prize and Audience Award at the 2024 inaugural Dubai AI Film Festival.

证据：Evidence: 完成 5 分钟完整短片《Critterz》，系首部获得美国版权局正式登记的 AI 短片。Completed the full 5-minute short film 'Critterz', recognized as the first AI short film registered with the U.S. Copyright Office.

证据：Evidence: 制作完成 22 分钟叙事长片《Echo Hunter》，系影史首部获得 SAG-AFTRA 工会认证的 AI 影片。Produced the 22-minute narrative film 'Echo Hunter', recognized as the first SAG-AFTRA certified AI film in cinema history.

证据：Evidence: Emmy 获奖导演运用全 AI Veo 2 流程制作 4 分钟 Wu-Tang Clan 官方无对白音乐录影带《Mandingo》。The Emmy-winning director directed the 4-minute official Wu-Tang Clan dialogue-free music video 'Mandingo' utilizing a full-AI Veo 2 pipeline.

证据：Evidence: 工作室依托固定虚拟演员 Bryan Stenler 进行 IP 运营，产出《Feed Me Lies》（2:02）与《Tonight We Rise》（2:08）等多部作品。The studio operates an IP built around persistent virtual actor Bryan Stenler, producing multiple works including 'Feed Me Lies' (2:02) and 'Tonight We Rise' (2:08).

证据：Evidence: 产出时长达 6 分 55 秒的英语对白 AI 动作喜剧系列首集《Couriers of Fortune》Ep.1〈Chain Letter〉。Produced the 6:55 English-dialogue AI action comedy series premiere 'Couriers of Fortune' Ep.1 'Chain Letter'.

证据：Evidence: 制作 5 分 33 秒作品《THE PUNCH BOWL》作为完整参赛片入选参与 Higgsfield Global Film Festival 2026。Produced the 5:33 work 'THE PUNCH BOWL', entered as a complete film festival competition entry in the Higgsfield Global Film Festival 2026.

证据：Evidence: 发布 28 秒大衣广告样片结合 ElevenLabs 配音，并在另一条目中专门对照同一酒馆场景的 Kling 4.0 同步生成声音与 Mirelo SFX 2.0 后配两版声音（两处声音实践成立标强）。Released a 28-second coat spec ad featuring ElevenLabs voiceover, and comparatively tested Kling 4.0 sync audio against Mirelo SFX 2.0 post-scored audio on the exact same tavern scene across two separate entries (sound verified across multiple works, marked strong).

证据：Evidence: 在 30 秒 9 镜头短片中实现人与狗形象全程一致，并开源 30 秒千禧年家用 DV 活人感方法（先定首帧与人物参考、拆动作时序、动作日常化并留真实时间）。Achieved full character consistency for human and dog across a 30-second 9-shot video, open-sourcing a methodology for millennial home DV realism that locks subject references and sequences everyday realistic actions (both character and action verified across multiple entries, marked strong).

证据：Evidence: 开源 10 秒搭 5 分钟电影的分镜框架并逐帧拆分摄影机角度参考，吉卜力风咖啡动画配环境音与音乐，且总结过 6 个网站包含 201 种动作音效的代码对齐配音流程。Shared a storyboard framework building 5-minute films with frame-by-frame camera angle references, paired Ghibli-style coffee scenes with soundscapes, and systematized a 6-site audio workflow aligning 201 action sound effects with visuals (sound and narrative verified across multiple entries, marked strong).

证据：Evidence: 接入 Renoise MCP 实现智能体全流程自动完成分镜规划、视频素材生成、剪辑、配乐到渲染出片。Integrated Renoise MCP into an agent pipeline that automates storyboard planning, clip generation, editing, scoring, and rendering into final deliverables.

证据：Evidence: 为冲浪动作短剧重新剪辑并加入新镜头与调色，并产出 Seedance 2.5 制作的日系手绘爱情动画《White Butterfly 白蝴蝶》。Re-edited the surf action drama with new shots and color grading, and produced the Seedance 2.5 Japanese hand-drawn romantic animation 'White Butterfly'.

证据：Evidence: 在 Dreamina 画布用人物三视图锁定主体一致性，通过连续分镜图每格生成 8 到 10 秒视频并在画布时间线内直接剪辑导出，还编写 CLI 脚本一键替换主角出片。Locked character consistency using 3-view sheets on Dreamina canvas, generated 8-10s video per storyboard node, edited directly on the canvas timeline, and scripted a CLI to swap characters across full pipelines.

证据：Evidence: 开源全流程文档，以 Blender 真实比例搭景锁定空间关系后经 Claude 生成分镜板与机位语言，并发布《RED RISING》三部曲科幻奇幻预告片。Open-sourced a complete workflow using true-scale Blender blockouts to lock spatial relationships before prompting Claude for storyboards and camera language, also producing the 'RED RISING' sci-fi trilogy trailer (narrative verified across multiple entries, marked strong).

证据：Evidence: 分享用群像人物图选角并转为详细角色卡锁定人脸一致性的流程，并发布公开剧情与分镜时序的 16 秒奇幻战斗短片《骑士小猫反杀恶龙》。Shared an AI casting workflow converting ensemble images into character sheets to lock facial consistency, and released the 16-second fantasy combat short 'Knight Kitten Defeats Dragon' with complete storyboard timing.

证据：Evidence: 发布超级英雄故事短片，探索直接给 AI 角色下达表演、声音与性格指令，强调独特视觉风格与角色一致性。Released a superhero story short directing AI character acting, voice, and personality, emphasizing distinct visual styling and character consistency.

证据：Evidence: 展示混合制片测试，将简单灰盒实拍素材通过 AI 流程转化为具有电影感的悬崖场景，以及将实拍演员喝咖啡转为完整酒吧场景。Demonstrated hybrid filmmaking pipelines transforming gray-box footage into cinematic cliff environments, and turning live-action coffee drinking into a fully-rendered bar setting (set verified across multiple entries, marked strong).

证据：Evidence: 利用大语言模型分析塔可夫斯基视听语言并执导成片，实测探讨大场景动态与快速追踪等复杂运镜稳定性。Analyzed Tarkovsky's cinematic visual language via LLM to direct short films, evaluating stability in rapid tracking and complex camera movements.

证据：Evidence: 生成修女与堕落天使角色表并在暗黑奇幻对峙中固定站位与洞窟地理，同时用单张参考图切分 15 秒音乐做口型同步短片，并公开 Suno 结合 Three.js 手指同步音乐视频做法。Built character sheets and locked geographical positions across a 2-part dark fantasy duel, segmented 15s audio with single reference images for lip-synced videos, and integrated Suno with Three.js for finger-synced music videos (character and sound verified across multiple entries, marked strong).

证据：Evidence: 发布 30 秒不切镜的一镜到底人类史短片（转场藏在穿过镜头的物体后），制作逆向毁灭移动边界长镜头，并在画布建立角色表锁定视觉形象后接续动作。Released a 30-second continuous one-take history of humanity hiding transitions behind objects passing the camera, produced reverse-destruction long takes, and locked character sheets on canvas across action sequences (camera and editing verified across multiple entries, marked strong).

证据：Evidence: 专门分享跟随动作的运镜提示词，完成 Kling 4.0 15 秒雨夜闹市自行车追逐一镜到底，以及 15 秒贴前轮并绕骑手肩膀的山路追车长镜头。Shared action-following camera prompts, producing a 15-second uncut rainy night bike chase on Kling 4.0 and an uncut motorcycle chase tracking wheels and shoulders on mountain roads (camera and action verified across multiple entries, marked strong).

证据：Evidence: 公开鱼眼短片四步法：先用灰色模特定好动作，再用 Kling 4.0 按秒数生成并按音乐调速重接，最后带音乐预览叠加角色与背景图合成正片。Shared a 4-step fisheye video method: blocking actions with gray models, generating clips on Kling 4.0 with music retiming, and compositing characters and backgrounds over preview audio.

证据：Evidence: 公开 15 秒竖屏 9:16 自拍 Vlog 喜剧提示词（含分段动作、镜头与反向词），并发布皮克斯风格 3D 动画家庭喜剧短片。Published a 15-second vertical 9:16 selfie Vlog comedy prompt with segmented actions and camera directions, also producing Pixar-style 3D animated domestic comedy scenes.

证据：Evidence: 在 CineMap 内预排车辆、编排动作、布置机位并反复调整角度，定稿后再生成电影感警匪追车戏序列。Pre-visualized vehicles, blocked actions, and staged camera angles in CineMap before generating cinematic police chase sequences.

证据：Evidence: 单人历时约 3 周完成短片全流程，借助智能体工作流进行剧本拆解、场景元素准备并推进至剪辑与后期。Completed a full short film solo in approximately three weeks, utilizing agent workflows for script breakdown, scene asset staging, editing, and post-production.

证据：Evidence: 展示《Lost Garden》第 1 集“Source Tree”场景，实现上方成片镜头与下方分镜帧在同一项目内严格对应。Demonstrated 'Lost Garden' Episode 1 'Source Tree' aligning final shots directly with corresponding storyboard frames within a unified project workspace.

证据：Evidence: 在《The Gunslinger》中将义体加速变速节奏与镜头衔接直接写进提示词一次直出，并用 Opus 5.5 剪辑 3×3 光剑网格音乐视频随音乐触发视觉延伸。Prompted speed-ramps and camera transitions directly into single-take prompts for 'The Gunslinger', and edited a 3x3 lightsaber grid music video triggering visual effects to musical cues.

证据：Evidence: 发布黑暗奇幻武打短片（盲眼武者红色沙漠对决、发丝化作触手并运用手持运镜），并发布写实 AI 视频公开提示词。Released a dark fantasy martial arts short featuring blind warriors, hair tentacle transformations, and handheld camera work, alongside photorealistic AI video studies.

证据：Evidence: 在 MV《ripple》中用 Claude 生成全尺预剪辑与离线编辑（成片七成沿用），结合 Kling 4.0 液体动态与 Suno 配乐。Created the music video 'ripple' by generating full-length offline pre-edits via Claude (retaining 70% in final cut), combined with Kling 4.0 fluid dynamics and Suno score.

证据：Evidence: 建立定格动画角色设计模板（规范面部、发型、服装面料、鞋履与姿态），并开源毛茸茸怪兽角色槽位替换与黑暗奇幻电影片段。Built stop-motion character design templates standardizing facial features, fabrics, shoes, and poses, also open-sourcing modular furry monster designs and dark fantasy scenes (character verified across multiple entries, marked strong).

证据：Evidence: 把动作戏拆解为 8 个节拍以精确控制出拳打斗节奏，并用三件套工具产出短片。Structured action sequences into 8 discrete beats for precise combat rhythm control in punching shots, producing shorts via multi-tool pipelines.

证据：Evidence: 发布 30 秒奇幻短片并公开分场景提示词，场景设定为雪夜松林中的僻静冬日木屋与微型奇幻生物。Released a 30-second fantasy short with scene-by-scene prompts, staging a secluded winter pine forest cabin and miniature fantasy creatures.

证据：Evidence: 建立镜头语法系列（旋转镜头、变焦推镜与 Snorricam），并在 Snorricam 系列中连续锁定医生与雨夜快递员面部状态以呈现人物心理。Created a cinematic grammar series spanning rotation, crash zoom, and Snorricam, locking character facial framing on doctors and rain-soaked couriers to convey emotional states (camera verified across multiple entries, marked strong).

证据：Evidence: 搭建全本地 MV 工作流 PopTok，整合 ACE-Step 配乐、LTX 转场 LoRA、InfiniteTalk 口型同步与 Whisper 词语对齐。Built the all-local MV workflow PopTok integrating ACE-Step scoring, LTX transition LoRAs, InfiniteTalk lip-sync, and Whisper word alignment.

证据：Evidence: 发布动作戏纯靠提示词实现高速运镜、场景破坏表现与斩击特效。Produced an action combat scene achieving camera movements, environmental destruction, and slashing VFX purely through prompt engineering.

证据：Evidence: 在 Google Antigravity 内利用 AI 完成绿幕去除与虚拟形象动画视频剪辑合成。Executed green-screen removal and avatar animation video editing/compositing within Google Antigravity.

证据：Evidence: 分享格斗戏做法：为每个角色与道具各做参考表锁定一致性，并将提示词按秒拆分为连续动作镜头清单。Shared combat scene methods: building reference sheets for characters and props to lock consistency, and sequencing prompts into second-by-second shot lists.

证据：Evidence: 资深剪辑师分享广告智能剪辑流程，按风格构建剪辑技能库并通过过往成片反馈循环打磨。Senior commercial editor shared advertising AI workflows, structuring skill libraries by editing style and refining via feedback loops.

证据：Evidence: 制作木帆船风暴短片，利用镜头拉远反转揭示风暴置于水杯内，并按海浪、风声与木头吱呀进行多轨分层音频处理。Created a wooden ship storm short with camera pull-back reveals into a tabletop glass, separating multitrack audio across waves, wind, and wood creaks.

证据：Evidence: 公开《绮月流霞》东方幻想角色提示词，详述品月色礼服配海棠红点缀与月白轻纱、环状云鬟发型配珐琅花片与水晶珠链等汉服妆造设定。Published character prompts for 'Qiyue Liuxia' in an Eastern fantasy style, detailing Hanfu costume and styling specifications including moonlight-blue robes with crabapple red accents, flowing gauze, looped cloud-hair buns, enamel hairpins, and crystal bead chains.

证据：Evidence: 分享 500 多集短剧角色一致性做法：用 FLUX 出正面、侧面、全身三张定妆图，在 ComfyUI 固定工作流中每镜头调用并配合 IP-Adapter 压制漂移。Shared character consistency methods for 500+ micro-drama episodes: generating front, side, and full-body lookbook stills via FLUX, locking them in ComfyUI templates for every shot, and using IP-Adapter to suppress drift.

证据：Evidence: 开发 AI 剪辑系统 Video Cut，内置 80+ B-roll 模板，支持从逐字稿到成片全流程自动预填分镜并导出到剪映细剪。Developed the Video Cut AI editing system featuring 80+ B-roll templates, supporting automated storyboard pre-filling from scripts to rough cuts with export to CapCut for fine editing.

证据：Evidence: 使用 Codex 搭配 Renoise MCP 分析参考视频完成 30 秒广告复刻，规划分镜并于云端环境完成素材剪辑、转场、渲染出片。Replicated a 30-second commercial spec using Codex and Renoise MCP, analyzing reference videos to plan storyboards and completing editing, transitions, and rendering in cloud environments.

证据：Evidence: 设计视频分镜提示词架构：要求角色卡外观固定不改，按普通日常+意外+收尾结构拆解 0–5 秒分秒时间轴，规划景别与动作。Structured video storyboard prompt templates that lock character sheet appearances and segment timelines into 5-second intervals with established setups, incidents, and punchlines.

证据：Evidence: 独立制作全 AI 短片《FLY》，以手工剪纸西洋镜风格完成全片，使用 Claude Code 撰写剧本、分镜、导演与剪辑，并整合 Seedance 2.5、GPT Image 与 ElevenLabs。Produced the all-AI short film 'FLY' in a handcrafted cut-paper diorama style, using Claude Code for scripting, storyboarding, directing, and editing alongside Seedance 2.5, GPT Image, and ElevenLabs.

证据：Evidence: 制作 25 秒电影感短片《The Art of Becoming – Amy Ghost》，在 Adobe Firefly 工作流中整合 AI 视频生成、时间线剪辑、素材混剪与调色。Created the 25-second cinematic film 'The Art of Becoming – Amy Ghost', integrating AI generation, timeline editing, stock footage mixing, and color grading within Adobe Firefly.

证据：Evidence: 制作 30 秒 90 年代伦敦 35mm 胶片质感短片，包含 8 个定时剪辑镜头、原生对白与真实街头音效，通过 4 格角色表与 6 视图锁定角色一致性。Produced a 30-second 1990s London 35mm film aesthetic short with eight timed cuts, native dialogue, and authentic street sounds, using 4-panel character sheets and 6-view boards for consistency.

证据：Evidence: 在 40 分钟内完成全流程 AI 历史短片制作，走通剧本、分镜、生成到 DaVinci Resolve 剪辑与修改成片。Produced an AI historical short in 40 minutes, executing an end-to-end workflow from script and storyboard through generation, DaVinci Resolve editing, and final revisions.

证据：Evidence: 在 Sumi Studio 中先把角色重绘为 4 张干净锚图，使后续人物镜头生成只参考锚图以锁定角色一致性。Repainted characters into four clean anchor images within Sumi Studio so that subsequent generation only referenced anchor stills to lock consistency.

证据：Evidence: 在 Kling 4 制作万圣节 MV 中运用 Ref2Video 工作流，以成片静帧替代角色表与地点参考图进行画面与主体锁定。Applied Ref2Video workflows while producing a Halloween MV on Kling 4, utilizing finished stills rather than character sheets and location boards to lock subjects.

证据：Evidence: 以射手、门将与球场三方参考图在 Pika H3 Flex 中分别生成两段各 15 秒视频，并通过后期剪辑拼成 30 秒连续一镜到底点球场景。Generated two 15-second clips using reference images of the shooter, goalkeeper, and pitch in Pika H3 Flex, then edited and spliced them into a seamless 30-second penalty kick one-take.

证据：Evidence: 用 Seedance 2.5 做 30 秒未来风汽车广告，公开完整 9 镜分镜提示词；另有首尔生活短片与超市变身短片，均公开完整提示词。Published full 9-shot storyboard prompts for a 30-second futuristic car commercial made with Seedance 2.5; two further shorts (a Seoul lifestyle piece and a supermarket transformation) also ship with complete prompts.

证据：Evidence: 围绕作品《MIRROR》将实拍影像自由排布版式并修改相机运动，结合 vista4D、meridian、Blender 与 Flame 等工具链重构运镜。For the work MIRROR, freely arranged live-action layouts and modified camera movements, restructuring camera language across a toolchain including vista4D, meridian, Blender, and Flame.

证据：Evidence: 使用 Three.js 进行视觉预演排布镜头，贯通 Krea2、Seedance 2.5、Suno 与自研 Claude 编辑工具完成万圣节魔女竞速短片分镜。Utilized Three.js to pre-visualize and stage shots, integrating Krea2, Seedance 2.5, Suno, and a proprietary Claude editing tool to structure the Halloween witch racing sequence.

证据：Evidence: 阿根廷 Purga Studio 四人团队（Manuel Montes de Oca、Bautista Vidal Iraola、Martin Montes de Oca、Joaquin Iraola）作品：高乔人的日常被一只总回到屋顶的羊反复打断；评审 Paul W. S. Anderson 称其「视觉、色彩、构图与声音都很强」，Alex Proyas 称「有趣、动人、富于巧思」；50 万美元头奖。By Argentina’s four-person Purga Studio team (Manuel Montes de Oca, Bautista Vidal Iraola, Martin Montes de Oca, Joaquin Iraola): a gaucho’s routine is repeatedly interrupted by a sheep that keeps returning to his roof. Juror Paul W. S. Anderson praised its “very strong visuals, color, composition and sound”; Alex Proyas called it “funny, moving, thoughtful and visually inventive.” $500,000 first place.

证据：Evidence: Zack Murray（美国）作品：汽车影院的头号影迷终于等到属于自己的任务；评审 Anderson 称「故事出色、执行到位」，Phedon Papamichael 称「概念好、执行好」；10 万美元第三名。By Zack Murray (United States): the drive-in’s biggest movie fan finally gets a mission of his own. Juror Paul W. S. Anderson called it a “terrific story, very well executed”; Phedon Papamichael: “Good concept, well executed.” $100,000 third place.

证据：Evidence: 整个世界由纸折成：一枚真实图钉给了一位焦虑的中学合唱团员「把一切固定在原地」的力量，纸面世界即舞台；荣誉奖 1 万美元。In a world folded from paper, a real pushpin gives an anxious middle-school vocalist the power to fix anything in place; the paper world itself is the stage. Honorable Mention, $10,000.

证据：Evidence: 手绘质感的战乱世界里，一个少年试图证明自己，爱与暴力同息——手绘场景世界是全片的主体表达；荣誉奖 1 万美元。A boy trying to prove himself, set in a hand-painted, war-torn world where love and violence share the same breath; the hand-painted world carries the film. Honorable Mention, $10,000.

证据：Evidence: 一次电梯之旅穿过九层越来越超现实的楼层，把乘客一步步推向看不见的清算——九段式结构叙事；荣誉奖 1 万美元。A man’s elevator ride spirals through nine increasingly surreal floors, each one pulling him closer to a reckoning he can’t see coming. Honorable Mention, $10,000.