Skip to content
Date and source

Dates follow the news timeline in Beijing time and include both endpoints.

Oct 15Thu
  1. Hugging Face Daily Papers24

    AVC-Reward: An Event-Centered Audio-Video Consistency Evaluation Framework

    Addressing multi-event combinations, multi-layer audio mixing, and asymmetric cross-modal relationships in joint audio-video generation models, researchers proposed the event-centered evaluation framework AVC-Reward, which decouples audio and video events at the event level and performs cross-modal cross-validation, aggregating into a coarse-to-fine three-level consistency metric.

  2. Hugging Face Daily Papers22

    Dusting the Phantom: allocation strategies for decaying content in dynamic Gaussian representations

    To address the problems of fast-moving small objects being easily lost and details in highly dynamic regions being hard to recover in dynamic Gaussian representations, the research proposes two strategies, Seed Injection and Temporal Biased Relocation, which use temporal appearance changes and reconstruction error to locate under-reconstructed regions, and use time-resolved optimization signals to guide Gaussian relocation, respectively. Without changing the underlying representation and while maintaining a fixed Gaussian budget, the two improve reconstruction quality on multiple dynamic scene datasets.

  3. Hugging Face Daily Papers22

    Mine, Sample, Render: test-time diversity alignment for generative models

    Researchers propose MINE-SAMPLE-RENDER (MSR), a test-time method that mines interpretable categorical patterns from a corpus of real samples and samples them statistically, then uses the generative model only as a conditional renderer to output the sampled configurations, thereby improving generation diversity. On tasks including weather report generation, user simulation, web design, text-to-image, and agent-generated 3D assets, MSR improved alignment with the corpus distribution and diversity of outputs.

  4. Reddit StableDiffusion · 创作者与工作流分享18

    How to Reproduce nano banana 2's Generation Results Locally?

    A user tried to reproduce nano banana 2's paid generation results locally using krea 2, flux 2 dev, flux 2 klein 9b, and qwen image edit 2.1, but none came close. Their need is to input two or more face reference images to generate a scene, then convert the result to video using h3 minimax. The video step works fine, but they hope the starting frame can be generated locally instead of paying for it.

  5. Reddit StableDiffusion · 创作者与工作流分享42

    Testing LTX-2.5 with Different Prompts in ComfyUI — Surprising Results

    A user tested LTX-2.5 in ComfyUI with five prompt sets, covering luxury perfume ads, AI influencer talking heads, sci-fi scenes, anime character animation, and first-to-last frame transitions. The first-to-last frame transition results were notably poor, and the author believes more testing is needed to judge LTX-2.5's performance in that workflow, asking other users for their experiences.

  6. Reddit StableDiffusion · 创作者与工作流分享20

    How to Make AI Videos Look Like Casual Phone Footage Instead of Cinematic Shots?

    A user generating "casual desktop product shots" videos with self-hosted MiniMax H3 found that no matter how they prompted, cinematic zooms and camera movements always appeared — even adding smartphone/handheld descriptions and trying with and without realism LoRA still resulted in overly polished output. They asked whether this natural phone footage effect mainly depends on the model, prompts, reference images, or settings, and hope to see a working workflow.

  7. Reddit StableDiffusion · 创作者与工作流分享12

    Reddit Discussion: Why Haven't the Latest Open-Source Models Arrived? Are Image/Music/3D/Audio Fields Stagnating?

    Reddit user ascot_major posted questioning whether open-source generative models are stagnating, stating that in recent months no major new models have emerged in key areas like image, music, 3D, and audio. The post lists H3 MiniMax as the last major video model, Krea as the last popular image model, Trellis 2 and Hunyuan 2.1 as the latest free 3D generation models, and Yue-2 as the latest music model, noting they still use Chatterbox-tts for audio.

  8. Curious Refuge film gallery22

    How Mike Staniforth's AI Short Film 'Creation of Adam I' Lets Viewers 'Fly Into' Michelangelo's Masterpiece

    Mike Staniforth's AI art short film 'Creation of Adam I' loops the camera from behind God's shoulder, sweeping across his outstretched arm and through the famous gap between their fingertips. The immersive perspective transforms Michelangelo's iconic composition into a spatial journey viewers can enter, no longer watching from the outside but moving through the painting itself.

  9. Curious Refuge film gallery22

    Chris Castor's AI Western Short Film 'The Savior'

    Chris Castor released the AI western short film 'The Savior,' telling the story of former outlaw Eli Mercer, who is tracked down by his old gang seven years after leaving. A young gunslinger watches from a distant ridge. Set in a harsh winter frontier, the story unfolds a heavy narrative about guilt, justice, and forgiveness amid tense standoffs and gunfire, questioning whether a person can escape their past or can only learn to carry it.

  10. Curious Refuge film gallery22

    Rubeen Raj's AI horror short film "White Field": Mongolian scouts hunt in a snowy forest but become the prey

    Rubeen Raj has released the AI horror short film "White Field," which tells the story of two Mongolian scout soldiers tracking a monster that slaughtered their camp in a dense winter forest, only to become the prey themselves. The film combines a historical setting with monster horror, creating sustained fear between duty and the instinct to retreat, as the hunt gradually turns into a desperate fight for survival.

  11. 我爱计算机视觉 · 公开公众号 RSS37

    NeurIPS 2026 | OmniMem enables long video generation to directly retrieve the complete history

    Northeastern University and Adobe Research have proposed OmniMem, achieving full-history long video generation for the first time and allowing the model to retrieve the memories it currently needs in real time. The method keeps full-resolution historical KVs on the CPU, recalls them to the GPU on demand, and customizes Triton and CUDA kernels around block-level retrieval.

Oct 14Wed
  1. note · singtothesky79 · 图像视频创作分享10

    Self-taught AI video creation ③: Using Gemini to make CANVA-generated AI characters move

    After generating AI characters in CANVA, the author tried having Gemini make a video using those characters, and successfully produced an animated clip of multiple characters in different colors playing. The author believes that if this workflow is further refined it could be used to make animation, but progress is currently stuck for the time being, and exploration will continue.

  2. note · kazu_fuwa · 图像视频创作分享76

    MiniMax H3 local deployment hands-on test: comparison of 7 configurations at 16GB VRAM, a 5-second video takes about 2 minutes 40 seconds

    The author ran MiniMax H3 with ComfyUI on a PC with 16GB VRAM and 32GB RAM. After comparing 7 configurations, a 5-second video was completed in as fast as about 2 minutes 40 seconds, while the official 20-step setting takes about 12 and a half minutes.

    Why it matters: The author tested 7 configurations on a machine with 16GB VRAM and provides reusable speed-up combinations and a record of pitfalls encountered.

  3. TelosCompass · note AI视频创作实践34

    How to keep a product's color and shape consistent across two scenes in AI video generation

    When producing AI video generation scene by scene, the prompt written for the previous scene is not automatically carried into the next scene, so the product's color and shape can easily look like a different item after splicing. The approach is to write the unchanging features (such as pale green color, rounded square shape, leaf-shaped engraving) as a "共通メモ" (shared memo) and reuse them verbatim in each scene's prompt, changing only the display range and action. Before generating the video, compare the two scenes' images side by side and check them against the real product, then decide whether to splice.

  4. note · toshia_fuji · 图像视频创作分享22

    Gemini 4 Argon launches globally, Google DeepMind releases its latest frontier model

    Google DeepMind's Gemini 4 Argon has launched globally, with priority access for Pro and Ultra subscribers. The model uses a sparse Mixture-of-Experts architecture, supports native multimodal input of text, images, and audio, has a context length of up to 1 million tokens, and is priced at $2 per 1 million input tokens and $10 per 1 million output tokens.

  5. 壹娱观察 · 搜狐公开专栏28

    AI film and TV companies may be castles in the air: Utopai's model, talent, and distribution challenges

    壹娱观察 questions the concept of "AI film and TV companies," citing Utopai as an example and pointing out that its video model Utopai X is post-trained on top of MiniMax H3, with its foundational capabilities coming from outside. Its flagship projects 《Cortés》 and 《The Most Serious Fart》 still rely on traditional key creatives such as Oscar-nominated screenwriters and Pixar veterans for endorsement. The article argues that AI has only entered the production process, while distribution and creative judgment still depend on traditional film and TV companies; AI short dramas have already replaced more than 90% of live-action short dramas but have not brought greater demand.

  6. open-ai-canvas · 开源创作工具发布18

    Open-source creation tool v1.6.8.3 released: fixes canvas cloud sync and image size configuration

    The open-source creation tool released v1.6.8.3, fixing the canvas cloud sync issue: after resource addresses are normalized, local persistence and sync state finalization are completed even without cloud write operations, and unconfirmed modifications are preserved during concurrent editing, viewport changes, and replay of old journals.

  7. Frank Houbre · AI电影制作实战31

    The 2030 Film Studio: AI Filmmaker Frank Houbre Walks Through It Room by Room

    Using his own Outerframe Studio as an example, AI filmmaker Frank Houbre walks through how film studios will change in 2030, room by room. He judges that physical soundstages will still exist by then, but will be smaller and rely more on rentals, with shooting focused on performance capture and environments left to generation; the new SAG-AFTRA contract (July 2026 to June 2030) and the transparency obligations of Article 50 of the AI Act have already locked in some of the rules.

  8. 逛逛GitHub66

    Noiz AI Open-Sources HelixWorld World Model, Capable of Real-Time Generation of Interactive Visuals and Sound

    Noiz AI has open-sourced the world model HelixWorld. After giving a starting image and filling in a scene description, you can move with WASD and rotate the view with the arrow keys, and the model continues generating subsequent visuals and sound based on the current scene and operations.

    Why it matters: The open-source world model generates visuals and spatial sound fields together, which can be used to understand the current deployment barriers and capabilities of real-time interactive world generation.

  9. note · fordtaurus116 · 图像视频创作分享22

    AZ8 Hands-On: Making a 30-Second Video from a Single Image on Infinite Canvas with Seedance 2.5

    AZ8's Infinite Canvas strings together image generation, video generation, and the AI agent Zumi on the same canvas; for video generation you can choose models such as Seedance 2.5. The author tested gradually expanding materials from a single image to make a 30-second video, with intermediate compositions and 15-second drafts all kept on the canvas. AZ8 also offers a Creator Program, where after applying and passing review you can get perks such as membership and production credits, and you can publish production services through the Marketplace.

  10. note · kt_kt_89 · 图像视频创作分享24

    After Handing Video Generation to an AI Agent, I Let Codex Take Over the Operator Role

    In addition to a local video generation agent, the author introduced Codex as a higher-level operator, letting it read logs, state, and Reviewer correction instructions, decide on sending back for revision and resuming from checkpoints, while the author only sets policy. This division of labor has been trialed since v0.5.64; finalized execution and safety state management remain local, and boundaries such as budget caps and retry limits are unchanged. The author emphasizes that this is a personal hands-on impression, not a quantified benefit, and that Codex itself also has usage and execution limits.

  11. 雷峰网22

    AIsphere's Hong Kong Office Officially Opens, Landing at Cyberport to Build a Hong Kong AI Hub

    AIsphere announced on October 13 that its Hong Kong office has officially opened at Hong Kong Cyberport, positioned as an AI hub serving local enterprises, government agencies, universities, and creators, and simultaneously launched the "Digital Smart Hong Kong · AIsphere AI Co-Creation Plan Across Industries." At the opening, the company showcased PixVerse R2, the world's first general-purpose real-time world model with scaling capability, and also signed multiple strategic cooperations with its domain innovation, Hong Kong Polytechnic University, and others. PixVerse currently covers 177 countries and regions worldwide, with cumulative registered users exceeding 170 million.

  12. Reddit StableDiffusion · 创作者与工作流分享36

    Reminder: raising CFG on distilled few-step video models makes output worse, not more faithful to the prompt

    On LTX-2.5, raising CFG from 1 to 4 caused image quality to collapse, with oversaturation and texture distortion. The reason is that the model is a guidance-distilled model, prompt adherence is already baked into the weights, and base inference runs only 3 denoising steps; raising CFG deviates from its operating point during distillation. The author reminds those migrating from the SDXL/SD1.5 habit of CFG 7 that they need a different mindset.

  13. Monster社区 · 中文ComfyUI实测与工作流22

    Playable on 8G! The must-install local 125B modified Strata: engine-level modifications, the Uncensor version supports native video analysis, and the integrated package and ComfyUI nodes are open-sourced

    The 125B modified version of Strata has been released. After engine-level modifications, it can run locally with just 8GB of VRAM, and the Uncensor version supports native video analysis. This version simultaneously open-sources an all-in-one package and ComfyUI nodes, accompanied by the T8 ComfyUI all-in-one package V10, the Zhenzhen all-in-one package V4.6B, and multiple trainer images.

  14. 智东西62

    China's top 3 video models go their separate ways: the diverging paths of Kling, Seedance, and Vidu

    Zhidongxi reviewed the path divergence among leading Chinese video model players around the National Day holiday: on September 28, Kling AI officially announced its next-generation video generation model Kling 4.0, and on October 8, Shengshu released the first preview version of its next-generation flagship model, Vidu Q4 Preview, starting at 0.09 yuan/second.

  15. cnBeta60

    Man uses AI video to mock AITO, Seres sues for 1 million yuan in damages

    Seres Auto has filed a lawsuit over a case of rumor-mongering via AI-generated video, seeking 1 million yuan in damages from the defendant. The case will be heard on October 23 at the People's Court of Liangjiang New Area, Chongqing. The male defendant is accused of using AI tools to fabricate footage of an AITO vehicle's body collapsing while driving, and, with added voiceover, smearing product safety before posting it on a short video platform. Seres has made four claims: deletion of the infringing content, a public written apology, 1 million yuan in damages, and bearing litigation costs; some legal professionals remind that AI is only a production tool and cannot serve as a reason for exemption from liability.