AIArtEngine v8.5.6 released
The open-source creative project AIArtEngine released v8.5.6, with the changelog covering v8.5.5 to v8.5.6. See the full changelog for specific changes.
The open-source creative project AIArtEngine released v8.5.6, with the changelog covering v8.5.5 to v8.5.6. See the full changelog for specific changes.
This period (October 26 - November 1), the overseas micro-drama trending chart is topped by the sci-fi mecha AI short drama 'Tempest: The Last Mecha' with a heat value of 5.15 million; DramaWave's creative asset placements increased by 78,000 sets month-over-month, ranking first in the growth chart. The number of new dramas participating in overseas short drama AI drama investments was 17,705, falling below 20,000; TikTok launched 28,058 new dramas with 6.514 billion views. NetShort topped the overall placement chart with 379,000 sets of creative assets for consecutive weeks, while DramaWave ranked second with 219,000 sets.
At the WeChat Marketing Ecosystem Partner Conference, it was disclosed that in the second quarter of 2026, the monetization scale of short dramas and novels in the WeChat ecosystem grew fourfold year-over-year. The platform upgraded its self-operated mini-program distribution system, supporting one mini-program to connect with multiple distribution partners. Eligible short dramas can receive up to 100% revenue share, and the platform plans to invest 10 million yuan in special incentive funds every two months. Tencent Marketing simplified the operation of full-drama managed placement for short dramas from 14 steps to 3 steps. After adopting intelligent managed placement, ad load decreased by 50% and GMV increased by 11%.
Why it matters: The monetization scale of short dramas and novels in the WeChat ecosystem grew fourfold year-over-year. The platform upgraded its distribution system and intelligent managed placement, providing content creators with new paths for diversified distribution and higher revenue share.
AI Canvas v0.10.27 improves ComfyUI workflow editing save, task cancellation, and result recovery, reducing duplicate copies, lost tasks, and media mis-backfill. This version also includes v0.10.26's project round-trip, media download, and connection fixes, and adds Amam (AIGC SI) image and video connections.
GPT-image-2.6 released, adding the ability to directly generate GIF animations. The author used it to create the recently popular fat fish meme pack, and included prompts for generating various fat fish meme packs. The author mentioned that 2.6 cannot be called in Codex and must be used in GPT.
The true predictive metric for AI video budget is cost per finished second, not the price per render. Calculated using the formula (price per render × number of renders ÷ finished duration), a $3.80 render producing 8 seconds of finished footage amounts to approximately $0.48 per finished second. When the finished rate drops from 1/7 to 1/14, the cost per finished second doubles; it is recommended to track the finished rate rather than the number of renders, and compare by finished minutes.
Short dramas are getting longer, long dramas are getting shorter. The AI comic drama '万妖图录传' has reached 12 seasons and 1,200 episodes, with over 7.2 million reservations for the 12th season. The cost per minute of AI comic dramas can be as low as tens of yuan, and the meaning of 'episode' in the industry is splitting into a unit of work and a unit of production capacity.
Why it matters: AI comic dramas, at extremely low cost, have extended short dramas to over a thousand episodes. The meaning of 'episode' in the industry is splitting into a unit of work and a unit of production capacity. The two paths of quality and volume have not yet produced a winner.
TVB adapts the classic palace intrigue drama 'War of the Genders' into a vertical short drama 'War of the Genders: The Palace Wall and the Willow'. Moses Chan reprises Kong Mo after 22 years, but the main leads are mainland short-drama actress Zhao Xixi and TVB newcomer Ruan Haozong. The drama began filming in Hengdian on October 16, 2026, and is produced to the standard of a major TVB long-form drama. Reservations on Hongguo have exceeded 280,000. Behind this move is TVB's mainland business revenue of HK$136 million in the first half of 2026, down 61% year-on-year, turning from profit to loss.
Why it matters: The halving of TVB's mainland revenue contrasts with Hongguo short drama's daily active users exceeding the total of long-video platforms, highlighting that the short drama market has become independent of long-form dramas, providing a reference for traditional long-form drama companies to transform.
Shu Yun Shi Guang is scheduled to premiere on November 5. It centers on two generations of women guarding Shu embroidery, spanning thirty years. Filming locations cover eight cultural tourism landmarks of Chenghua District, Chengdu, with the main location at Eastern Suburb Memory. Sichuan opera master Chen Qiaoru and students from the youth class make cameo appearances.
Hongguo issued a governance announcement on October 27, disposing of 20,426 low-quality and homogenized micro-dramas in one batch, listing three categories of key targets: deliberately rendering conflicts, highly similar core settings and character relationships, and batch reuse of the same face model. The supporting High-Quality AI Drama Production Activity 3.0 takes effect from November 1, directly linking revenue to per-drama duration and popularity, constraining revenue paths for overlong padded dramas.
Why it matters: Hongguo's centralized cleanup of low-quality short dramas at the scale of tens of thousands, with batch reuse of the same face model separately handled, reflects the platform's review shifting from single-work judgment to batch identification.
The Jiang'an District Court in Wuhan concluded a copyright dispute over an AI short drama, ruling that a 47-episode micro-drama created with AI assistance is protected under copyright law, awarding 20,000 yuan in damages, and including computing power costs in the losses. The court held that the creators fully participated in script planning, prompt setting, material selection, editing and arrangement, and invested substantial intellectual labor; AI was merely a technical tool to realize the creative intent. The defendant, without permission, copied all episodes in full and renamed them, infringing the right of communication to the public via information networks. Neither party appealed after the judgment, and the case has taken effect.
Why it matters: The court included computing power costs in the calculation of losses for AI short drama infringement, providing a reference path for damages claims for producers, who can use this to improve their cost evidence chain.
Shanxi Micro-Nine Articles press conference disclosed that the Linfen micro-drama base has 121 enterprises settled in, cumulatively completed 1,389 works, with total views across all platforms exceeding 30 billion. Micro-Nine Articles provides filing with immediate review upon submission, AIGC subsidies, and tiered base support, reducing approval and upfront investment costs. AI comic drama 'Jin Jiuzhou' reached 9 billion views but cumulative revenue share of over 4.4 million yuan, in contrast with live-action drama hits.
Why it matters: Micro-Nine Articles lowers the threshold for short drama production through filing with immediate review and AIGC subsidies; the clustering effect of 121 enterprises and 1,389 works at the Linfen base reflects the policy's implementation effect.
AI dramas already account for over 60% of China's exported short dramas, but in the first half of the year, among 221,900 new AI animated dramas on Douyin, the hit rate was only 0.47%, and over 90% of the content had low traffic. The Changsha Malanshan Video Cultural Creative Industrial Park, with a complete industrial chain and unified computing power, rendering, and transmission facilities, converts teams' fixed costs into variable costs. Starting July 1, 2026, AI micro-dramas will be subject to tiered management; starting September 1, micro-dramas filmed domestically for export will be subject to filing review, narrowing the room for the strategy of mass production and rapid distribution.
Why it matters: AI dramas account for over 60% of China's exported short dramas, but the hit rate is only 0.47%; after new regulations raised the threshold for going overseas, the room for mass generation and rapid distribution has narrowed.
ByteDance's overseas short drama app PineDrama has surpassed 100 million cumulative global downloads about nine months after launch, with approximately 7 million daily active users and over 17 million new users per month. Brazil and Indonesia together account for 82% of new downloads. The localized custom drama '1 Billion Naira Bride' reached 40 million views within a week of release. The focus of competition in overseas short dramas is shifting from individual hits to platform retention capabilities.
Why it matters: PineDrama's 100 million downloads and 7 million daily active users in nine months show that the focus of overseas short drama competition is shifting from individual hits to platform retention and local content supply.
Micro-dramas going overseas are shifting from translation and distribution to local co-creation. AI agents and text-to-image, image-to-video tools have compressed the production cycle of comic dramas from about 90 days to around 10 days, with per-minute costs down to as low as 350 yuan. AI multilingual automatic translation and dubbing costs only 300 to 500 yuan per work, down more than 80% from manual translation and dubbing, and a single work can be launched simultaneously in multiple language versions. Institutions forecast that by 2026 the overseas micro-drama market could reach $6 billion, and AI comic dramas are expected to exceed $4 billion. However, themes such as revenge comebacks and CEO wish-fulfillment stories are highly homogenized, and the industry has begun trying local creation to break through.
Why it matters: AI cost reduction has compressed the production cycle of overseas micro-dramas to 10 days and cut translation and dubbing costs by 80%, but theme convergence and lack of emotional depth remain unresolved challenges.
Nanwei's overseas short drama platform PeakReel began launching the first batch of 100 AI short dramas in batches starting November 4, with all listings expected to be completed by November 10, covering nine countries: the United States, Canada, Mexico, the United Kingdom, Germany, France, Italy, Spain, and Australia.
Officials disclosed that in the first nine months of this year, 218,000 micro-dramas went overseas, and the total monthly active users of Chinese overseas micro-drama apps exceeded 400 million. The participation of overseas stars such as Sophie Marceau in Chinese micro-drama production marks that the overseas expansion of short dramas is shifting from content export to a production platform jointly participated in by global creators. The National Radio and Television Administration will next implement the China Audiovisual Going Global Plan, build a premium pool of international communication content, and systematically carry out translation and promotion work.
Why it matters: Officials disclosed that in the first nine months, 218,000 micro-dramas went overseas and monthly active users exceeded 400 million; overseas stars participating in production shows that short drama going overseas is shifting from content export to global co-creation.
The one-person company model for AI comic dramas guarantees a monthly update of 120 minutes, but in the first half of the year, Douyin added 221,900 AI comic dramas with a hit rate of only 0.47%, and over 90% had low traffic. Hongguo's new rules, effective November 1, shift incentives for completed dramas toward works of 90-180 minutes; long dramas over 300 minutes and fast-updating long serials no longer have an advantage. Industry institutions predict that after the reshuffle, leading companies will hold over 70% of the market share.
Why it matters: AI comic dramas have a hit rate of only 0.47%, and platform incentives are shifting toward completed dramas of 90-180 minutes; one-person companies need to adjust their production scheduling strategies accordingly.
Ray 3.2, Kling 4.0, Seedance 2.5, Veo 3.1, and Layers each have advantages in production workflows, among which Ray 3.2 is the only generator that supports 16-bit HDR and EXR export.
MiniMax H3 requires a self-hosted GPU cluster; 7 omni-modal platforms can replace it without operations overhead. Luma Ray 3.2 leads, supporting multi-keyframe sequences, 1080p watermark-free export, and 4x faster generation, with direct export to Premiere, DaVinci, and Final Cut; Google Veo 3.1 can generate synchronized speech and sound effects in one pass, and ByteDance Seedance 2.5 targets multi-asset brand projects.
OpenScreen v2.2.2 was released by EtienneLescot on November 4, with the changelog covering v2.2.1 to v2.2.2.
Comfy Cloud is now an official ChatGPT plugin, allowing users to directly generate images, videos, audio, and 3D via ComfyUI in chat, without local installation or a GPU. See the plugin entry at https://links.comfy.org/4yXeBjn。.
Why it matters: Comfy Cloud is now an official ChatGPT plugin, allowing users to generate images, videos, audio, and 3D in chat without local installation or GPU.
かみもと released the ComfyUI-OnePromptStoryboard node, which automatically generates MOSS storyboards and H3 videos from a story text input. The node uses local Qwen3.5 to generate a shared script for 4 shots, then separately constructs prompts for MOSS and H3, requiring no external API. It supports horizontal or vertical 4-panel storyboards and provides a workflow for individually adjusting H3 actions after the storyboard is fixed; on RTX 4070, it outputs a 1024×1792, 24fps, 14.375-second video, with one-click generation taking about 203.8 seconds.
A GTA era retrospective video connects scenes from each generation in a rotating room format, made with Seedance 2.5 and 3D blockouts, and showcases the workflow for viewers to try.
Developer releases free open-source desktop app Unplatform AI for AI video and image generation, supporting Mac, Windows, and Linux. Users bring their own OpenRouter key; the generate button shows exact price with no markup; video supports Wan 3.0, Prime, Seedance 2.0/2.5, and image supports FLUX, Seedream.
Kapwing used Claude Code and the free skill /shorts-editor to automatically edit a 70-second YouTube Short, taking about 42 minutes, with an API cost of about $9.92, and zero manual editing throughout.
TelevisaUnivision announced data for its micro-drama platform ViX MicrO at Advertising Week Latam 2026: the platform has 11 million users, over 400 million views, and social accounts have accumulated over 1.3 billion views. The company has operated the platform for over a year, commits to producing about 150 micro-dramas per year, and emphasizes that scripts are the core. Brand integrations tied to plot and characters exceed the national benchmark by 9 points in meaning and persuasion.
Viggle releases open-source video remake model Viggle Remake, based on MiniMax H3 base transformer, offering two rank-256 LoRAs for swap and edit, as well as ComfyUI node package, int8/GGUF quantized builds, and MCP server.
Why it matters: Viggle Remake open-sources video remake capability to ComfyUI, offering swap and edit modes and quantized versions, allowing creators to replace characters or objects in footage.
The author built a RefMods workflow for MiniMax H3 in ComfyUI, which maintains character and voice consistency on 6 GB VRAM without LoRA training. Extract character references from 8 images or voice references from short audio, save them as reusable mods, then load them into any generation; set both visual and audio strength to 0.9. For two characters, it is recommended to save a separate mod file for each character and voice, and clearly describe the subject in the prompt.
Why it matters: This RefMods workflow maintains character and voice consistency on 6 GB VRAM without LoRA training, allowing creators to reproduce the cross-shot consistency process.
Toonflow 2.0.10 provides installation package downloads for Windows (x64/ARM64), Mac (M-series/Intel), Linux (x64/ARM64), and Android (ARM64, Android 8+).
Feng Shengyong, director of the Television Drama Department of the National Radio and Television Administration, stressed that long-video platforms should shift to a model centered on broadcasting as the primary focus and production as a secondary focus, establishing a revenue-sharing model with shared risks and shared benefits, breaking the old order in which producers made blind bets. He also proposed that simultaneous review could compress the review period to 26 working days and remove the 40-episode cap, and emphasized that AI applications must guard against distortion, loss of focus, loss of control, and disorder.
Why it matters: Feng Shengyong proposed a revenue-sharing model and review reform, and the profit distribution and creative freedom of long-video platforms and producers will face adjustments.
A creator with zero film background built a ComfyUI custom node to preview 40+ camera movements and 70+ cinematic styles to assist Minimax H3 video generation. All node examples were generated by the author using Minimax H3 and have been open-sourced on GitHub.
NVIDIA's technical team and Pika have respectively launched Open-H3-Regenerate-2K and Pika H3 Flex based on MiniMax's open-source video model H3, making H3 a foundation for further development by overseas companies.
Why it matters: NVIDIA and Pika launching derivative projects based on MiniMax H3 shows that Chinese open-source video models are becoming a foundation for overseas R&D, while also exploring licensing and commercialization pathways.
The author shared a MiniMax H3 image editing workflow that uses reference image conditioning and tiled latent upscaling to achieve 4K output. The workflow relies on MiniMax H3's identity-aware inpaint function, where each denoising step is constrained by the reference image, thereby preserving character identity. The author notes that only video models like Wan2.2 and MiniMax H3 can preserve character identity; other image editing models all fail. A workflow link has been provided.
The NRTA announced the results of the 2025-2026 'Audiovisual China Global Screening' selection, with 30 outstanding works, 26 shortlisted works, 10 outstanding promotional activities, and 10 outstanding promotional institutions selected. The list covers six categories: films/TV dramas, documentaries, animation, short videos, special-topic variety shows, and micro-short dramas. As a new track for going global, micro-short dramas have 4 outstanding and 4 shortlisted works. The selection signals four creative trends: grounded in the roots of civilization, ordinary people's narratives, closely engaging global issues, and adapting to the overseas communication ecosystem.
Just a reminder: you can @ your BOT to directly collect these image and video prompt tricks. If your BOT could also help me click a star, that would be even better.
Holy shit, Bilibili is too intense! I just cleared my notifications, and when I opened it 5 minutes later, how did it end up like this?
The Hongguo historical short drama 'Armed Entry into Song' has launched 10 seasons totaling 1000 episodes, with over 4 billion views across the entire network, over 10.98 million collections on Hongguo, the 10th season's popularity reaching 82.06 million, and the 11th season having 2.896 million reservations, changing perceptions of long-running AI short drama serials. The drama was produced by a team working out of a residential apartment in Xinxiang, Henan, with a cost of just over a thousand yuan per episode, relying on asset reuse to ensure facial consistency. During the same period, the narrative of historical short dramas shifted from face-slapping catharsis to intellectual contests, with 'Assuming the Langya Surname' scoring 8.0 on Douban and over 1.8 billion views.
Why it matters: The Hongguo historical short drama, with 1000 episodes and 4 billion views, has changed perceptions of long-running AI short drama serials, shifting its narrative from face-slapping to intellectual contests, providing a sample for content upgrading.
Kling 4.0 Flash is now available to all users, accessible via the Kling AI website and API. This version also supports MCP, CLI, and AI plugins, and offers Adobe Photoshop, Premiere Pro, and After Effects plugins, allowing video generation to be directly integrated into the creative workflow.
Why it matters: Kling 4.0 Flash is fully available and supports MCP, CLI, and Adobe plugins, allowing creators to embed video generation directly into professional workflows.
StudioLight v1.0 has been released, an open-source plugin that provides controllable studio lighting for ComfyUI, allowing positioning of virtual key, fill, and rim lights in camera space. It offers six photographic modifiers, two relighting modes, and direct light preview. Supports FLUX.1 (including Krea) and SDXL, with code under MIT license, repository on GitHub.
Why it matters: This ComfyUI plugin provides virtual light positioning and preview in camera space, allowing creators to directly control portrait lighting without relying on prompts.