Skip to content
Date and source

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

Oct 23Fri
  1. 机器之心SOTA模型 · 公开公众号 RSS38

    Canva releases the Craft image generation model; Shanghai AI Lab open-sources the ShuSheng-A1.5 research agent

    Canva released the Craft image generation model on October 22. It uses a two-stage process—composition planning first, then pixel rendering—supports 2K output, more than 30 aspect ratios, and in-image text in 55 languages, and is available through Canva Dream Lab and a public API.

  2. Reddit StableDiffusion · 创作者与工作流分享22

    MOSS LoRA: one-click generation of realistic Instagram-influencer-style photos

    A LoRA called MOSS has been released that directly produces Instagram-influencer lifestyle photos with warm tones, soft colors, and natural lighting, requiring no extra filters or retouching. The LoRA is compatible with Moss base, SFT, and Kromos, suits everyday scenarios such as selfies, casual phone snapshots, street fashion, cafés, and travel, and is recommended at a weight of 0.6-1.0 with no trigger word needed.

  3. MindStudio Blog · 创作分享42

    Recraft V4.1 Flash Released: ~1.3-Second Image Generation with Shareable Custom Styles

    Recraft V4.1 Flash leads with fast image generation at approximately 1.3 seconds per image, at the cost of first generating at a lower resolution, then optionally upscaling and optimizing quality through the enhance feature. Custom styles support building from reference images, built-in example prompt previews, and newly added sharing via link for others to apply. Recraft still offers a free plan, allowing users to try V4.1 Flash and style tools.

  4. MindStudio Blog · 创作分享75

    LTX 2.5 Open Weights Released: 22B Video Model Available for Self-Download and Operation

    Lightricks released LTX 2.5 open weights on Hugging Face. This is a 22B-parameter diffusion model that can be downloaded and run on your own GPU. It also provides a full dev version and a distilled accelerated version, along with bf16, int8, and nvfp4 quantization formats.

    Why it matters: Lightricks open-sourced a 22B video model along with audio and multi-shot capabilities. The article breaks down the components of the download package and the self-hosting requirements.

  5. MindStudio Blog · 创作分享78

    LTX 2.5 Explained: An Open-Weight Video Model Supporting Synchronized Audio and Native Multi-Shot

    LTX 2.5 is an open-weight AI video model that produces audio and video in the same generation pass and supports multiple coherent shots within a single generation. The model is available in standard and distilled versions; weights can be downloaded, called via a hosted API, or deployed on your own GPU. The official team says the new diffusion video decoder improves performance on faces, textures, on-screen text, and fast motion. The article also documents the process of self-hosting on a remote Nvidia GPU and testing lip-sync and facial consistency with dialogue shots.

    Why it matters: The article breaks down LTX 2.5's three features—audio-video co-generation, native multi-shot, and self-hostability—to help assess the practical boundaries of open-weight video models.

  6. SuSu_酥酥58

    Domestic 3D Model Hi3D V3.0 Released, V1.0 Open-Sourced Simultaneously

    Domestic AI 3D model Hi3D V3.0 has been officially released. The author compared it using the same character image and stated that it is closer to the original in terms of outline, volume, and local details than GPT-6 Astra. V3.0 features commercially usable 2048³ precision and textures up to 8K; V1.0 is open-sourced based on Sparc3D, with training data expanded to approximately 2 million, a 90% slice success rate, and support for MCP integration with Blender and game engines.

  7. Reddit StableDiffusion · 创作者与工作流分享60

    PixText Released: A Small, Fast Model for Fixing Text in Generated Images

    The author released PixText, a small model specifically designed to fix garbled text and Chinese characters in generated images, with approximately 0.95B parameters. The model is trained directly in pixel space to avoid VAE compression losses on small text details, requires no text encoder or VAE, and uses MeanFlow to complete text repair in just 4 sampling steps. The author also provided ComfyUI nodes, ready-to-use workflows, code, and model weights.

  8. IT之家71

    Volcano Engine releases the Seedance 2.5 4K model, aimed at film-grade image generation

    ByteDance's Volcano Engine has released Seedance 2.5 4K, calling it the strongest model in the Seedance 2.5 generation, with specialized training to learn character expressions, subtle movements and image details. The model supports a Draft sample mode, allowing users to first verify storyboards and polish shots with 480P samples before generating the 4K final cut, and it features 10-bit color depth; it is currently available only for trial by some enterprise customers, and some film and television production and post-production companies at home and abroad have already tried it.

    Why it matters: Seedance 2.5 4K splits sample verification and the 4K final cut into two stages, allowing film and television teams to judge how to trade off image quality and cost accordingly.

  9. 可灵 Kling72

    Kling 4.0 launches on idea.ai

    Kling 4.0 is now available on idea.ai. According to idea.ai, this version can generate a 30-second scene from a single prompt, maintain character consistency across shots, and support switching between live-action, anime, and 3D styles.

    Why it matters: Kling 4.0 launches on idea.ai, supporting 30-second scene generation from a single prompt and cross-shot character consistency, and can be evaluated against existing output pipelines.

Oct 22Thu
  1. 量子位 · 搜狐公开专栏75

    Moss Intelligence's open-source MOSS-Image surpasses 1.3 million overseas downloads in 4 days

    Moss Intelligence's open-source 5B-parameter text-to-image model MOSS-Image gained 1.3 million downloads in overseas communities within 4 days of launch, received 1.4k Stars on GitHub, and topped the Hugging Face text-to-image trending list.

    Why it matters: The 5B-parameter open-source model gained 1.3 million downloads in 4 days of launch; the community has already produced accelerated and quantized versions and LoRAs, allowing observation of its actual performance in text rendering and layout control.

  2. 可灵 Kling62

    Kling officially teases the Kling 4 model

    The official Kling account posted a brief teaser with the hashtags #KlingAI #Kling4, without disclosing specific capabilities or a launch date. The quoted content shows that Kling 4.0 is already available on OpenArt, supporting first/last frame and intermediate frame guidance, up to 10 reference images, text-to-video or frame-to-video, and video editing, with 50% off Pro and Wonder plans before November 20.

  3. 可灵 Kling75

    Kling 4.0 launches on OnSolo

    Kling 4.0 is now available on OnSolo, supporting native 30-second video generation in a single pass while maintaining motion stability in fast action and large camera movements. This version offers stereo sound and more precise lip-sync, supports a 21:9 widescreen aspect ratio, and allows up to 15 references and 10 keyframes to be stacked per task.

    Why it matters: Kling 4.0 launches on OnSolo, offering parameters such as native 30-second video per generation, 21:9 aspect ratio, and stacking of multiple reference images, making it easier for creators to evaluate against their existing workflows.

  4. 可灵 Kling72

    Kling 4.0 launches on Adobe Firefly

    Kling announced that Kling 4.0 is now available in Adobe Firefly, and users can try it directly via a link provided by Adobe. This version is integrated into Firefly under the name Kling 4.0, and the official announcement did not disclose further feature details.

    Why it matters: Kling 4.0 is integrated into Adobe Firefly, letting creators learn about the entry point for this version within Adobe workflows.

  5. comfyorg · YouTube 创作视频78

    ComfyUI launches FLUX 3 Max, supporting image-to-video and text-to-video workflows

    ComfyUI launched FLUX 3 Max, a latency-optimized version of FLUX 3 Video that can generate 10-second 720/768p clips with native audio in about 12 seconds. The official release also provides links to two workflows, image-to-video and text-to-video, and lists use cases such as comparing multiple candidates, previewing motion while tuning parameters, batch-producing ad variants, and keyframe-guided storyboards.

    Why it matters: ComfyUI integrates FLUX 3 Max, offering a latency-optimized version that produces 10-second 720/768p clips with native audio in about 12 seconds, along with entry points for image-to-video and text-to-video workflows.

  6. 可灵 Kling72

    Kling teases the Kling 4 video generation model

    Kling's official account teased the next-generation video generation model with #Kling4, without giving specific parameters in the post. The Morphic release information it cited says Kling 4.0 supports native 30-second generation, up to 10 keyframes, 15 references, on-screen text consistency, and multiple languages, and offers 4K and 10-bit HDR, with a 10% discount for generations on Morphic before November 20.

    Why it matters: Kling officially teases Kling 4; cited information indicates it supports native 30-second generation and 4K 10-bit HDR, which can be used to gauge its long-shot output capability.

Oct 21Wed
  1. 可灵 Kling72

    Kling 4.0 launches on the fal platform

    Kling 4.0 is now live on fal. The official description calls it a new-generation native multimodal video model with improvements in complex motion stability, character and scene reference consistency, and cinematic narrative control. Currently, the model is only available through fal UX and fal Agent, with API access coming soon.

    Why it matters: Kling 4.0 launches on fal, with the official announcement citing three capability changes: complex motion stability, character and scene reference consistency, and narrative control.

  2. 可灵 Kling72

    Kling 4.0 launches on Higgsfield

    Kling 4.0 is now live on Higgsfield, supporting Omni Reference to place characters, objects, and audio in the same generation, and can generate a 30-second scene in a single pass. This version also supports controlling each beat with start, middle, and end frames, with native 4K labeled as coming soon.

    Same story, the pick is 《Kling 4.0 launches on Higgsfield: supports 30-second generation and 10 keyframes》

  3. 骨朵 · 网易号19

    Tonight's Livestream Preview | Full Hands-On Test of Kling 4.0! Using Top AI Technology to Recreate the Authentic Ancient Sanxingdui Civilization

    Guduo Media will livestream a hands-on test of Kling 4.0 tonight, focusing on four upgraded features: 30-second native long video generation, complex camera movement, precise keyframe controllability, and all-purpose material reference replication. The livestream will be hosted by Guduo Media partner Mi Mumu, with film art director Infinite Canvas and screenwriter Koi Xiaojinjin breaking down two original Sanxingdui scripts from the dual perspectives of film art and script narrative, recreating the sacrificial site and the iconic bronze sacred tree scenes, and comparing ancient civilizations such as Chaoge, Luoyang, Loulan, and Atlantis.

  4. ComfyUI Wiki · 工作流与开源创作分享75

    Tencent Hunyuan open-sources interactive world model Hy-Nomas, supporting segmented generation from text and camera trajectories

    Tencent Hunyuan open-sourced the interactive world model Hy-Nomas on October 21, 2026, releasing weights, inference code, and training and data annotation code, along with an online demo and the Hy-Nomas-Bench benchmark.

    Why it matters: Tencent Hunyuan open-sources an interactive world model in which text and camera trajectories can drive long video generation segment by segment, with a benchmark and single-GPU configuration.

  5. ComfyUI Wiki · 工作流与开源创作分享81

    FastVideo releases FastH3 Omni: a reference-conditioned variant of MiniMax H3 that generates video with sound in 8 steps

    FastVideo released FastH3 Omni, a reference-conditioned variant of MiniMax H3; a prompt plus ordered image, video and audio references can generate clips with synchronized sound, requiring only 8 transformer evaluations instead of 50 for base H3.

    Why it matters: FastH3 Omni unifies reference images, video and audio into a single generation pipeline, and provides an item-by-item time comparison with base H3 to help assess local deployment costs.

  6. Reddit comfyui · 创作与工作流分享78

    Hao AI Lab open-sources FastH3 OmniRef, with ComfyUI support on day one

    The FastVideo team at Hao AI Lab has open-sourced FastH3 OmniRef, which generates video with synchronized audio from a prompt plus reference images, short video clips, or sound. The model needs only 8 steps instead of 50, is up to 11.9 times faster than the base MiniMax-H3, has entered the ComfyUI 0.42.2 workflow templates, and has open weights and training code.

    Why it matters: FastH3 OmniRef is released as open weights, produces video in 8 steps, and supports day-one ComfyUI integration, so creators can use it to evaluate local reference-to-video pipelines.

  7. FastVideo 视频模型 · 模型发布80

    FastVideo releases ComfyUI version model weights for FastH3 Omni

    FastVideo released ComfyUI-usable weights for the final version of FastH3 Omni, providing two diffusion models, BF16 and INT8 ConvRot, which can generate video and synchronized audio from prompts and ordered image, video, and audio references.

    Why it matters: FastH3 Omni releases ComfyUI-usable BF16 and INT8 weights and provides the way to connect reference images, video, and audio, making it easier for creators to evaluate local video production pipelines.

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

    MOSS-Image generates stunning movie posters: clear text, specifiable title position, 5B parameters and locally runnable

    When used to generate movie posters, MOSS-Image produces clear, readable text and can also specify the positions of the title, tagline, and cast and crew credits, with realistic layouts in sci-fi and horror styles. The model has only 5B parameters and can run locally.

Oct 20Tue
  1. note · toshia_fuji · 图像视频创作分享32

    FLUX 3 Max video model released, with MiniMax M3.1 Flash, DeepInfer and others updated around the same time

    fal Research and Black Forest Labs jointly released the FLUX 3 Max video model, which tied for first place in an independent third-party image-to-video evaluation, generating a 10-second 720/768p video in about 12 seconds, 11.5 times faster than the original FLUX 3 and with a quality improvement of 50 Elo.

  2. IT之家68

    Ant Ling opens source Ming-3.0 speech series models, covering speech synthesis, audio understanding and dialogue, and speech recognition

    Ant Ling has open-sourced the Ming-3.0 speech series, including three series—speech synthesis, audio understanding and dialogue, and speech recognition—with five models in total, which developers can select by task or combine into workflows. Ming-3.0-tts's Voice Design supports describing the target timbre in text, and can also set the timbre with reference audio; director mode can specify how lines are performed, emotional changes, stressed words, and pacing.

    Why it matters: Ming-3.0 splits speech synthesis, audio understanding, and recognition into five models, among which director mode can describe the performance style of lines in text.