Higgsfield发布了Short Drama Studio工具,输入一个剧本即可输出整季短剧剧集,能保持同一剧组角色在各集间一致,并允许为每个项目选择不同风格(如黑暗奇幻或90年代高中风)。
输出支持竖屏、宽屏和经典4:3比例,同时该工作室可通过Higgsfield的MCP接入,让Agent驱动整个生成流程而无需手动拖拽时间线上的片段。其发布预告片在一天内获得近50万浏览量,针对当前竖屏短剧热潮。
Higgsfield推出Short Drama Studio,支持输入剧本生成整季竖屏短剧,保持角色跨集一致,并可通过MCP由Agent驱动整个流程。
🎨 AI ART & VISUALS ROUNDUP — November 17, 2026
1️⃣ A SCRIPT GOES IN, A WHOLE SEASON COMES OUT
Higgsfield shipped Short Drama Studio: you feed it a script and it returns a full season of short drama episodes, holding the same cast consistent from one episode to the next and letting you pick the style per project, from dark fantasy to nineties high school. Output comes in vertical, widescreen and classic four by three, and the studio is also reachable through Higgsfield's MCP, so an agent can drive the pipeline end to end instead of you dragging clips around a timeline. The launch reel ran close to half a million views in a day, aimed squarely at the vertical drama boom.
🔹 @higgsfield_ai
2️⃣ KLING PREVIEWS ITS NEXT STILL IMAGE MODEL
Kling posted a reel of still images with three claims attached, stunningly realistic, sharp and precise, and one sentence: Kling Image 4.0 is coming soon. No date, no specifications, no benchmark sheet, which reads like a staged rollout rather than a finished release. It is still a useful signal, because the image side of this field has been the quiet half of the race while video models took every headline. The teaser cleared ninety thousand views in a day.
🔹 @Kling_ai
3️⃣ SEEDANCE 2.5 FLASH ARRIVES IN COMFYUI
ComfyUI added Seedance 2.5 Flash through Partner Nodes, and it is positioned as the workhorse rather than the flagship: it is said to beat Seedance 2.0 in output quality while costing less than Seedance 2.5. Text to video, first and last frame control and multi reference input are all there, and existing reference clips can be edited or extended instead of regenerated. Lengths run from four to thirty seconds at 720p with native audio, which is what makes it usable for previz, social cuts and anything that needs sound in the same pass.
🔹 @ComfyUI
4️⃣ BACKGROUND REMOVAL THAT RESOLVES HAIR AND FUR
Bria's RMBG 3.0 now runs inside ComfyUI through the same Partner Nodes system. The claim list is unusually specific: cleaner edges on hair, fur and motion blur, a true eight bit alpha matte rather than a hard binary mask, identical output for identical input on every run, and training done only on licensed data. Deterministic results are the boring feature nobody markets, and they are exactly what a compositing pipeline needs when a cutout has to survive three more edits downstream.
🔹 @ComfyUI
5️⃣ A BENCHMARK THAT GRADES PRICE, NOT JUST QUALITY
Vivix published its read of an independent comparison from One2X Datalab covering seven video models across twelve tasks. W1 was scored best value, with pricing starting around half a cent per second, placed alone in the top tier for physics understanding, and it shared a perfect score on the liquid physics test with Wan 3.0, with separate notes on camera work, crowd handling and multi shot continuity. The company's own demo is a fifteen second watch commercial built from a few reference images, priced at a handful of dollars for a hundred variations. For small studios the buying decision is value per second, not peak quality.
🔹 @VivixLabs_HQ
6️⃣ A MUSEUM GIVES ITS OPENING SLOT TO AN AI WORK
Pérez Art Museum Miami switches on Techno Firma on Saturday, a piece by John Fitzgerald with Godfrey Reggio, the filmmaker behind Koyaanisqatsi, which returns rebuilt as a living obelisk that regenerates itself. It is the museum's first artist driven AI exhibition, timed to Miami Art Week and running through the season. The framing matters more than the gadget: the institution is not adding an AI corner as a sidebar, it is handing the flagship opening of its art week programme to a machine assisted work.
🔹 @CatalogueART
7️⃣ TWENTY ARTISTS ASK WHO THE AUTHOR ACTUALLY IS
Museum Kranenburgh in Bergen, Netherlands, opened AI-Artist, gathered by guest curator Edo Dijksterhuis, bringing together more than twenty artists who treat AI as a smart tool while pushing back against the automation and authorship language around it. It stays on view until May 9, 2027, a strikingly long run for a show about a technology that will look different twice before the walls come down. That is the real story here: museums are committing exhibition space for years to a tool that is still changing underneath them.
🔹 @CatalogueART
💭 The pattern this week is that the big tools stopped selling raw generation and started selling formats: a season instead of a clip, an extended cut instead of a fresh render, a clean matte instead of a mask. Meanwhile two museums in Miami and Bergen committed real gallery space to work made with these systems, one for a week of art openings, one for the next eighteen months. The gap between what a model can produce and what a small production can pay for keeps closing, and the honest part of the conversation has moved to fractions of a cent per second.
Would you hand a whole season to one instruction and supervise the result, or keep every single shot under manual control 👇
#AIArt #GenerativeAI #AIVideo
来源:AlexAImaginator · x.com