Dhar Mann Studios, Fox Entertainment, and Holywater have scheduled the first five micro-dramas from their 40-title slate. The first, 'The Brothers Who Love Me,' will premiere on My Drama on November 25, with subscribers able to watch early on November 18.
Producer Will Packer launches micro-drama studio June, developing, producing, financing, and acquiring vertical video series aimed at Black audiences. NBCUniversal and HarbourView are investing, with the investment amount undisclosed. Will Packer serves as CEO, and his son Dom Packer serves as president. HarbourView will focus on brand partnerships. The first slate is being developed with Black women at its core, and titles, distribution platforms, or release dates have not yet been announced.
The China Association of Automobile Manufacturers will conduct research and evaluation on the payment status of suppliers to 17 key automakers, including FAW, SAIC, Dongfeng Motor, BYD, NIO, Xpeng, Li Auto, and Xiaomi, focusing on checking whether the payment terms for small and medium-sized enterprises do not exceed 60 days. Sources confirmed that Bu Sinan, vice president of 360 and head of its marketing department, was taken away by police for investigation last week; 360 has not yet responded. Affected by the decline in international gold prices, the per-gram price of domestic brand gold jewelry continued to fall on November 16, with the listed price per gram more than 530 yuan below the year's high.
A ComfyUI video novice user is looking for a video generation model suitable for AI influencers, with a 20GB GPU, previously using Krea 2 and Qwen Image 2.1. The user posted in the community asking for advice, saying they are not yet familiar with video production.
ComfyUI Workflow Templates v0.11.110 has been released to PyPI and can be installed via pip install comfyui-workflow-templates==0.11.110. This update is the 2026-11-17 template update, submitted by @comfyui-wiki in PR #1429, with changes covering v0.11.109...v0.11.110.
A user asked what good anime LoRAs are available in Moss-Image and reported that existing solutions are not effective. The user said that the sekali_p style has poor prompt adherence in scenes with more than three people, and using anime tags directly results in insufficient detail.
Seedance 2.5 Flash is now available in ComfyUI via Partner Nodes, supporting text-to-video, first/last frame, and multi-reference generation, with the ability to edit or extend reference videos, generating clips of 4-30 seconds, up to 720p, with native audio, covering aspect ratios from 9:16 to 21:9. The model outperforms Seedance 2.0 and costs less than Seedance 2.5.
Why it matters: Seedance 2.5 Flash strikes a balance between performance and cost, outperforming Seedance 2.0 and costing less than Seedance 2.5, providing a new option for teams that need rapid iteration.
Adobe launches Premiere AI Assistant, which can use natural language prompts to organize footage, assemble rough cuts, and optimize visuals and sound. The assistant consumes Firefly generation credits and currently supports tasks such as organizing projects, preparing media, assembling edits, and optimizing visuals and sound. It is available for trial in the Premiere public beta.
The author used a single RTX 3090 24GB, swapping VRAM in and out, to run a local agent and ComfyUI image/video generation simultaneously. Measured: LLM about 72 tok/s, SDXL image about 32 seconds, Wan 2.2 video 4-6 minutes. Data comes from 205 real requests and 35 rendering tasks, not synthetic benchmarks. The author concludes that 24GB can only accommodate either the LLM or one render at a time, and recommends using parallel 1 instead of 4.
The author tested RelayVSR 2x and 4x video super-resolution in ComfyUI. On an RTX 4070 Ti SUPER, a 5-second 0.5MP video took about 28 seconds at 2x and about 53 seconds at 4x.
Why it matters: The author tested the performance and quality differences between RelayVSR 2x and 4x in ComfyUI, and provides timing data on a mid-range GPU as a reference for choosing video super-resolution settings.
At the Adobe Max conference, Adobe announced new agents and tools for Firefly, Creative Cloud, and GenStudio, including Creative Agent, MCP integration, and Premiere generative media tools. The company also deepened partnerships with Google and Microsoft and showcased super-resolution tools from the Topaz Labs acquisition.
Why it matters: Adobe launches multiple AI tools in Creative Cloud and Premiere and opens MCP integration, allowing creators to evaluate the impact of new features on existing workflows.
aTwist has reached a partnership with Snapchat. Starting December 1, Snap's original short drama series will be available on aTwist, with mutual promotion between the two parties. The first batch of series includes Dead of Night, Breakwater, Endless Summer, Death Hacks, Denton Death Date, Road Trippin, Phone Swap, and Coming Out.
MOSS Image Users attempt to use reference images as visual input via the CLIP text encoder, with the model using Qwen 3.5 9B as a prompt enhancer; this method occasionally reproduces elements such as clothing from the input image, but is mostly completely ineffective. Users are seeking a method for consistent implementation.
The author built and tested a MiniMax H3 TTM motion tweening workflow in ComfyUI; generation took 250 seconds and GPU memory usage was 71GB. The workflow requires CUDA 13 and downloading multiple model files to specified directories, using SAM3.1 to specify moving objects and paths. The author reports that there are two controllers, which are difficult to adjust.
The author released a MiniMax-H3 V2V Masterclass tutorial. The workflows cover video editing, time-window and spatial chunking nodes, inpainting cropping, latent-space super-resolution, batch processing, and ripple editing, among others.
Kapwing published a Sheru Bhai AI video production tutorial, using Claude Skill to generate character reference images and a complete video prompt, then generating a 15-second vertical video with dialogue in one pass in Seedance 2.5. The tutorial also includes environment reference images, a second-by-second storyboard for an auction scene, AI photo generation methods, and a troubleshooting checklist for character consistency issues.
Why is my Krea 2 Turbo LORA so weak? A user trained a Krea 2 Turbo amateur photo style LORA using the Civitai trainer, but the style effect is very weak at 1.0 weight and needs to be pushed to 1.5-2.0 to be noticeable. The training used 90 image/caption pairs and a single trigger tag; among 4000 steps, the 3000-step checkpoint gave the best results. Although high weights can achieve the target style, they introduce AI-induced roughness, and the user hopes to seek improvement suggestions before sharing.
Automotive photography company EVOX sues ByteDance, accusing the TikTok parent company of copying its 7,095 images to train the Seedream AI image generator. EVOX says it has a car photo library accumulated over 30 years, covering thousands of car models.
Long-form video platforms such as iQIYI, Mango TV, and Tencent Video are competing for AI creators through three channels: competitions, universities, and short videos, and have formed three cooperation models: custom production, joining production teams, and revenue sharing. Under the revenue-sharing model, the production cost of AI short dramas with 30 or more episodes is about 70,000–80,000 yuan, and production quotes generally range from 1,000 to 2,000 yuan per minute. Different models correspond to different capital investment, creative autonomy, and revenue distribution.
The author tested MiniMax-H3-Image on a PC with 16GB VRAM, verifying that it can generate magazine covers with Japanese titles and subtitles, as well as character illustrations. Compared with P-H3-Image, the files are larger, but the Japanese text is more accurate; however, the hiragana 'ぱ' becomes 'ば', and the background sometimes turns black. A 4096×4096 image takes about 38 seconds to generate, but Pixubus does not currently support large-size output.
Long-video platforms such as iQiyi, Mango TV, and Tencent Video are recruiting AI creators through three channels: competitions, short videos, and universities, and have formed three cooperation models: custom production, joining production teams, and revenue sharing. iQiyi has signed 8 AI creation labels and 3 individual creators; Mango TV has signed or reached cooperation intentions with over a dozen AI creators; Tencent Video cooperates on AI short dramas through revenue sharing. Quotes for AI drama production generally range from 1,000 to 5,000 yuan per minute. Under the revenue-sharing model, production costs for AI short dramas with more than 30 episodes exceed 70,000 to 80,000 yuan.
Why it matters: The article reviews the three paths—competitions, short videos, and universities—used by iQiyi, Youku, Tencent Video, and Mango TV to recruit AI creators, as well as the three cooperation models—custom production, joining production teams, and revenue sharing—providing a reference for creators to evaluate platform cooperation.
Suno Bridge lets you drag AI music creations directly into a DAW timeline and automatically matches the project's tempo, key, and style. The standalone app runs in parallel with the DAW and supports VST and plugins. It is currently in beta and only compatible with Ableton Live 12 on Mac. Suno plans to expand compatibility.
The author shares a workflow for maintaining character consistency in ComfyUI: split the character description and scene description into two frozen blocks; the character block is never edited, while the scene block is changed for each shot. A fixed seed makes the armor texture and hair strands match. The recommended IPAdapter FaceID weight is 0.75-0.85; above 0.9 the character looks pasted on, and below 0.6 the face drifts.
Why it matters: The author shares a workflow that splits character and scene descriptions into two frozen blocks; combined with IPAdapter FaceID weight and a fixed seed, it can reliably reproduce character consistency across scenes.
Kuku TV partners with Flipkart to introduce SuperCoins rewards into micro-drama subscriptions, with users able to receive up to 100% of the subscription value back. Subscriptions made through the Flipkart SuperCoins Rewards Store are eligible for rewards, while subscriptions made directly through the Kuku TV app can use existing SuperCoins to offset up to 30% of the subscription fee.
Training subject LoRA with Minimax H3 Image yields poor similarity, while the same dataset performs excellently on Qwen 2.1, Krea 2, and Moss. Increasing training epochs or steps showed no improvement; the Minimax H3 video model actually learns subjects better on static images. Motion learning is also difficult, and the poster is seeking advice on training settings for high-fidelity similarity.
Minimax H3 on a 4090 achieves best quality with settings of 1322x768, 24fps, 20 sampling steps, without turbo and without LoRA; the duration is usually about 4 seconds. For upscaling to 2k, you can use Regenerate 2k (Regen2k IC just released) or minimax_h3_latent_upscaler_3d_conv_v1_fp16. The author tested the API at 1080p and believes the quality improvement is not worth the price.
Globo brings over 30,000 hours of content and 200 development projects to MIP Cancun, positioning Brazil as a global producer of quality content. Its key works include 'Quien Ama Cuida' and 'Tres Gracias', among others. 'Tres Gracias' reached over 146 million viewers in Brazil, while 'Quien Ama Cuida' impacted over 119 million viewers in its first 12 weeks. The company is also expanding into micro-dramas and vertical content, exploring international co-productions and new consumption models.
China Film Reservation Screening will launch a limited-time party for 'The Illustrated Legend of Ten Thousand Demons: The Stranger from Yunzhou' in cinemas in more than 200 cities nationwide starting November 20. The event is centered on the IP of the hit AI comic drama 'The Illustrated Legend of Ten Thousand Demons', launching a side-story storyline in which Jiang Yuechu accidentally enters a cyber cultivation alternate world. Audiences can purchase sessions through the China Film Movie Pass reservation screening section. The event also collaborates with the limited-edition cinema theme 'pain building', converting online IP popularity into offline consumption.
WaveSpeedAI's Dora shares an AI video character consistency shot workflow, using character reference kits, fixed identity details, and two-pass review to avoid character drift. Vidu 2.0 accepts 1-3 reference images, MiniMax H3 Highspeed accepts up to 9 reference images and 3 reference videos; when fixing drift, change only one input at a time.
Why it matters: This workflow provides specific practices for character reference kits, fixed identity details, and two-pass review, and can be directly transferred to multi-shot AI video production.
Shishi Technology releases a multimodal model service that employs adaptive acceleration technology based on historical cache residuals, achieving an end-to-end average 2.48x speedup while maintaining generation quality. The service also incorporates SGLang-optimized caching and memory management, and modality-adaptive token compression achieves a 1.33x speedup in the WorldSense benchmark, while the Context Engine bridges material understanding and video generation. The service covers API access, task orchestration, model routing, inference execution, and result delivery, providing support for large-scale AI video production.
MiniMax H3 has poor lip-sync performance, but by using the LTX audio node to pack audio into the latent before generating video, you can get LTX-level lip-sync with H3 image quality. The workflow is simple: just add the LTX audio input node and merge the audio and visual latents. The author provides a workflow download and uses a VHS filter to process the video into an 80s VHS tape style.
An ANIMA LoRA user, when generating multi-character scenes with ComfyUI and 5080, experienced crosstalk in character appearances, affecting only the main character and one other, while the rest remained default. They posted seeking advice on tools, prompts, LoRA, nodes, or plugins, and mentioned that Pony's multi-character prompt tool performed poorly.
ByteDance Volcano Engine announced that the Seedance 2.5 4K API is now fully available, and developers and enterprise customers can call this 4K video generation model through the interface. The model supports Draft (sample) mode, where end users can first use 480P samples to quickly validate storyboards, repeatedly refine shots, and generate the 4K final film once they are satisfied; it also features 10-bit color depth, making the image quality more suitable for the big screen.
Why it matters: With the Seedance 2.5 4K API now open, developers can directly call the 4K generation capability, and Draft mode provides a low-cost path for storyboard refinement.
Hu Yong, professor at Peking University's School of Journalism and Communication, wrote an article analyzing the industrial transformation of AI short dramas, pointing out that micro-short dramas are undergoing two profound transformations: from paid viewing to free traffic models, and from live-action filming to AI generation.
Hongguo short dramas launches AI virtual image rights confirmation capability; copyright holders can submit ownership materials for AI characters in the drama, and after review approval, a rights confirmation record is formed. The rights confirmation is linked with the pre-check tool launched in August; when pre-check determines portrait rights infringement, one can proceed to rights confirmation; after confirmation, characters can join Hongguo to open virtual character accounts, and top comic drama characters have already started livestreaming.
Why it matters: Hongguo provides rights confirmation and onboarding channels for AI drama characters, turning characters from one-time materials into long-term operable assets, offering copyright holders a new monetization path.
AI comic drama creator Xu Fei, on the Hongguo platform, with an average production cost of two to three thousand yuan per work, achieved over 100 million views and nearly a million yuan in revenue share for a single work. He chose urban scripts of the rational type for the lower-tier market, used AI tools for batch production, and emphasized that editing and narrative pacing are more important than visual quality. Hongguo's new rules from November favor completed dramas of 90-180 minutes; the platform has 21.66 million daily active users and 411 million monthly active users.
Why it matters: A zero-foundation creator used AI comic drama on the Hongguo platform to obtain nearly a million yuan in revenue share; their cost control and topic selection strategy can serve as a reference for similar creators.
Aamir Khan's Aamir Khan Productions and Indian film and television company Excel Entertainment announced a joint launch of a short-form drama platform, and brought in a major telecom operator as a distribution partner. The platform will integrate both parties' existing film and TV IPs, original short-form drama content, and works from digital creator collaborations. The specific name and launch time have not been announced. Redseer predicts that India's short-form drama market size will grow from INR 23 billion in FY2026 to INR 250 billion in FY2032.
The third installment of the costume short drama 'After Being Honored Through Her Son, She Is Doted On by the Royal Family' launched on Hongguo on November 13, with first-day views reaching 59.42 million, ranking second on the live-action drama trending list. The controversy over the male lead change did not dampen its popularity; the drama received a 9-point rating, 181,000 favorites, and 279,000 likes, with its heat value approaching 680,000 and pre-launch reservations reaching 4.71 million.