Skip to content
Date and source

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

Sep 23Wed
  1. openshorts · 开源创作项目发布28

    OpenShorts v2.0.0-alpha.26 released: MCP server enables AI agent topic-in, finished-video-out

    OpenShorts released v2.0.0-alpha.26, its first npm release, adding an MCP server that lets AI agents go directly from topic to finished video. This version also fixes the issue of keeping the best draft when scripts are too short, switching to expanding or trimming existing drafts instead, unifies visual review candidate numbering (3B vision model improved from 1/5 to 4/5), and supports optional local Ollama vision models. Documentation notes local model tiers: 7B can run, but 14B or above is recommended for release.

  2. note · lpp · 图像视频创作分享36

    YuE2 AI music generation cover unified GUI released, supporting RX9070XT and LoRA

    YuE2 launched a cover unified GUI that lets users drag cover source audio into the interface, automatically converting it to ABC Score notation and extracting lyrics, then generating cover music with one click. This version includes LoRA support and adds two updates: separating audio from video as a cover source, and integrating with JIZURA to auto-fill lyrics. On RX9070XT (ROCm10.x), 20 seconds of music can be generated in about 10 seconds.

  3. ArcReel · 公开创作项目发布28

    ArcReel v0.31.0 released: integration with custom ComfyUI workflows and endpoint marketplace

    ArcReel v0.31.0 supports importing custom ComfyUI workflows to generate images and videos, automatically identifying key parameters and allowing request preview and connection testing before saving. The settings page adds an endpoint marketplace where users can browse, install, update, and uninstall endpoints; prompt templates support viewing all templates and 36 art styles by workflow. After script confirmation, users can directly add or remove storyboards and modify narration on the timeline, and prompt generation only processes modified storyboards.

  4. ArchiGen AI · 图像生成与ComfyUI实战22

    How to evaluate AI rendering tools: Workflow testing from first success to second success

    ArchiGen proposes evaluating AI rendering tools by "time to second success" rather than the best single image shown by vendors. The test requires first writing down five project-specific acceptance criteria, then changing the viewpoint or design version and handing it to a second operator, while recording the number of attempts, operation time, wait time, and discarded outputs. The article also suggests documenting item by item whether prompts, settings, reference images, etc. can be saved or exported with the project, to determine whether the workflow is truly transferable.

  5. ArchiGen AI · 图像生成与ComfyUI实战49

    ComfyUI architectural rendering repair workflow: First globally enhance the entire frame, then use masks to fix vegetation, reflections, and material defects item by item

    ArchiGen proposes a ComfyUI architectural rendering repair workflow: first use a single global enhancement to establish exposure, color, and material baseline, freeze the approved full frame, then use masks to locally repair named defects such as vegetation, reflections, materials, and shadows item by item.

  6. Loomance · 本地AI动画实测34

    How to turn artwork into animation locally: A hands-on comparison of Loomance and ComfyUI on WAN 2.2

    On an RTX 5070 12GB GPU, both ComfyUI and Loomance can run the WAN 2.2 image-to-video model entirely locally, without cloud rendering or per-use fees. ComfyUI offers node-level control but requires manual configuration, while Loomance wraps the same model into three preset workflows: Stable Scene, First → Last Frame, and Wild Mode, allowing direct rendering from children's hand-drawn sketches as source images.

  7. Wireflow · AI短剧与视频制作教程31

    How to build AI workflow software for agencies

    Building AI workflow software for agencies centers on converting a single repeatable client deliverable into a flowchart that runs the same way every time. The guide outlines six steps: first select one deliverable and break it down into steps, then choose a build layer among custom code, general automation platforms (n8n, Make, Zapier), and visual AI canvases with APIs, where visual canvases can produce a usable version in as little as 1 to 3 days.

  8. Wireflow · AI短剧与视频制作教程66

    How to Create AI Song Ads on a Single Wireflow Canvas Without CapCut

    Wireflow publishes a tutorial on how to complete an 80-second AI music ad on a single canvas — from script, song, and word-level timeline to stills, motion, subtitles, and end credits — without ever exporting to CapCut or Premiere.

    Why it matters: A complete breakdown of the eight steps to produce an 80-second AI music ad, from writing lyrics to rendering, along with the actual cost of approximately 6,100 credits and a record of pitfalls encountered.

Sep 22Tue
  1. Krea Reference · ComfyUI创作项目发布58

    ComfyUI Krea Reference Node Updated to Version V11

    The Krea Reference node for ComfyUI has released V11, adding reference balance preservation, content-aware caching, combined slider budgets, temporal and preset controls, and smoother sub-value adjustments. Existing V9/V10/V1 workflows and recipe defaults remain unchanged, and portable examples are included along with notes on quality limitations.

  2. 小智博客 · 音视频实践60

    Hands-on Environment Tuning for Running MiniMax-H3 on ComfyUI with a Single Hygon K100_AI Card

    The author documents the hands-on process of tuning and running MiniMax-H3 on ComfyUI with a single Hygon DCU K100_AI card. MiniMax-H3 is an open-source 33B omni-modal video generation model by MiniMax, supporting three modes: text-to-video, image-to-video, and reference-to-video. It natively outputs 2K/15-second videos with stereo audio. Hygon's official solution recommends 144GB of VRAM per card.

  3. 毒眸 · 搜狐公开专栏60

    Orca Entertainment Launches AI Film and TV Production and Management Platform Whale AI, with Closed Beta Starting in October

    Orca Entertainment unveiled its AI film and TV production and management platform, Whale AI, at the 'AI + Culture and Media' forum of the 2026 Yunqi Conference. A closed beta is planned for October, with the first batch of slots reserved for professional film and TV creators and production teams. The platform organizes scenes, character performance, camera, virtual studio, and lighting adjustments into five precision control suites, offering 40+ camera parameters, 250+ combined scene editing options, 49 basic expression controls, and 100+ character pose placeholders. It also supports multi-role collaboration for large teams and project consumption tracking.

  4. ArchiGen AI · 图像生成与ComfyUI实战30

    How to Mark Locked, Adjustable, and Replaceable Parts in Architectural Renderings Before Your First ComfyUI Enhancement

    Before your first ComfyUI enhancement, attach a one-page protection brief to each architectural rendering, dividing the image into three zones — locked, adjustable, and replaceable — and listing verifiable architectural facts. Locked zones preserve roof edges, window openings, columns, stairs, property boundaries, and camera position. Adjustable zones are limited to materials, exposure, sky, and glass tint. Replaceable zones regenerate vegetation, people, and distant backgrounds according to boundary rules. Run only one operation at a time, and keep a comparison table of the source image and each checkpoint.

Sep 21Mon
  1. openshorts · 开源创作项目发布12

    OpenShorts v2.0.0-alpha.25 Released

    OpenShorts released v2.0.0-alpha.25, fixing an issue where newly saved keys in the interface now take effect without a restart, and resolving eight silent failures on Windows where AO internal modules were uniformly converted to file:// URLs. New features include local Ollama script writing (0 key), copying diagnostic information, and page routing tests. In English contexts, only OpenShorts is used (with Kaipian as a supplement).

  2. The Creative Possible · 创作教程与项目分享12

    Curious Refuge Launches AI for Micro-Dramas Course on Micro-Short Drama Creation

    Curious Refuge launched the AI for Micro-Dramas course, teaching creators to build vertical micro-short dramas from scratch using AI tools. Content covers series bible and pilot script, world-building and AI visual consistency, AI video and agent workflows, episode editing and sound design, professional pitch decks, and distribution and sales. The first cohort opens for enrollment on September 30 at 11 a.m. Pacific Time, with a beta discount of $50, lasting one week. Participants will complete an original series bible, pilot episode, and sellable pitch.

  3. ArchiGen AI · 图像生成与ComfyUI实战13

    Product Pages, Ranking Comparisons, Community Reports, and Project Field Tests Are Four Different Types of Evidence: Please Keep the Labels

    ArchiGen proposes using an "evidence ladder" to distinguish four types of AI rendering tool information: product pages are official claims, ranking comparisons are vendor or media roundups, community reports are user experiences, and project field tests are the only self-verification. Taking today's search as an example, Chaos compared six products by BIM integration, pricing, and use cases; Gendo listed itself as the best in its own ranking; and Maverick Frame compared ten products by BIM compatibility, control, speed, and pricing.

  4. ArchiGen AI · 图像生成与ComfyUI实战26

    Why ComfyUI Workflows Need to Be Maintained Like Software Products

    ComfyUI workflows are free to download, but keeping them running across personnel, projects, models, nodes, and version updates is maintenance work. The article suggests treating workflows as small software products packaged with six items: the graph, a dependency list, model locations, sample inputs, expected checkpoints, and a runbook, and accounting for five categories of costs: environment maintenance, graph maintenance, project adaptation, quality control, and operator support. The author also recommends establishing fixed test scenarios, verifying architectural facts such as opening counts, mullion rhythm, and roof edges before and after updates, and scheduling updates outside of delivery deadlines.

  5. Wireflow · AI短剧与视频制作教程18

    Best Video Creative Software for Agencies in 2026: A Roundup of 8 Tools

    Wireflow is rated the best choice for multi-model AI video pipelines for agencies, with its node-based canvas connecting script, image, video, and voice models to produce multiple versions of client deliverables in a single run. The list also includes Adobe Premiere Pro (post-production and broadcast delivery), Runway (generative shots), Descript (voiceover and podcast repurposing), Canva (client-editable brand templates), VEED, CapCut, and Creatify.

Sep 20Sun
  1. 王若风 · 开源创作项目分享62

    Qwen Open-Sources Qwen-Image-2.1: 7B Unified Text-to-Image and Image Editing with ComfyUI Support

    The Qwen team open-sourced Qwen-Image-2.1, unifying text-to-image and image editing into a single 7B parameter checkpoint, with native support for 2K generation and RGBA transparent output with alpha channel, up to 10 reference images, and local editing via circle, doodle, or independent mask selection.

  2. GitHub:AI视频开源项目发现62

    Yizhi Zhuozi Open-Sources Table-VideoSOP: Voiceover-Driven AI Video Production Workflow and Prompts

    Author Yizhi Zhuozi open-sourced Table-VideoSOP on GitHub, a set of SOPs, rules, templates, and prompts for voiceover-based videos such as science explainers, opinion pieces, and product demos, progressing through nine steps: initialization, DESIGN, script and voiceover, macro table, segment design, visual confirmation, assembly, sound and delivery, and cover publishing.

  3. Binyuan Multi Image Compare · ComfyUI创作项目发布10

    Binyuan Multi Image Compare V1.5.2 Fixes Missing Chinese Manual in Release Package

    Binyuan Multi Image Compare released V1.5.2, fixing an issue where Python decoded git output using cp936 during packaging, causing `使用说明.txt` to be silently skipped. The release package now includes all 8 files. Other content remains consistent with 1.5.1, including the "Generation Info" auto-panel (which traces upstream sampler, LoRA, model, and other commit parameters and overlays them on images) and simplified preview badges to input/frame numbers.

  4. Binyuan Sampler V6.5 · ComfyUI创作项目发布22

    Binyuan Sampler V6.5 Released: New External Sigmas, LoRA Settings, and Denoised Latent Output

    Binyuan Sampler V6.5 has been released with the version number unified to 6.5.0. It adds an "External Sigmas" input to support external sigma scheduling and custom sampling, and introduces a "LoRA Settings" input that can stack with the node's internal LoRA list. This version also adds a "Denoised Latent" output that returns x0 for downstream reuse; when both external positive and negative conditions are connected, CLIP loading can be skipped.

  5. ComfyUI-Easy-Media · 开源创作项目发布22

    ComfyUI-Easy-Media v1.3.2 Released: New MiniMax H3 Progressive SelfLift Sampling

    ComfyUI-Easy-Media released v1.3.2, replacing character context exchange with pure Drift-Control AV and adding progressive SelfLift sampling for MiniMax H3. This version also improves the multi-track editor UI and universal tiling, and fixes issues with video reference frame and audio-locked video frame alignment, context latent selection, and duplicate semantic tags in prompts.

Sep 19Sat
  1. 宝玉的分享 · 创作与项目分享62

    Baoyu uses GPT 6 Astra and Codex to create a 3D version of "The Peach Blossom Spring" and publishes the prompts

    Baoyu used GPT 6 Astra in ChatGPT Pro to generate a 3D web version of "The Peach Blossom Spring," splitting the full text into 20 scenes and supporting following the text into the scene, slow chapter-by-chapter reading, and switching between the original and vernacular text. He replaced the system TTS with real-person recitation by CCTV's Li Lihong, used SRT timestamps to sync chapter transitions and sentence highlighting, then had Codex split the Base64-embedded MP3 into separate files and replace the stream-and-bird ambience, shrinking the HTML from 5.6 MB to about 777 KB.

  2. ArchiGen AI · 图像生成与ComfyUI实战38

    A two-input controlled test in ComfyUI using line art and clay passes from a locked camera

    The tutorial proposes exporting two passes, line art and clay/simple material, from the same locked camera in ComfyUI, and using a ControlNet two-condition controlled test to see which input protects which architectural facts. The method fixes the checkpoint, VAE, sampler, seed and resolution, changing only the conditioning path, and runs four cases: clay alone, line art alone, equal-weight pairing and tuned pairing, with four seeds per case.

  3. Wireflow · AI短剧与视频制作教程22

    Best batch image generation tools of 2026: a comparison of 8 including Wireflow, OpenArt and Recraft

    2026 batch image generation tools fall into three categories: batch prompt apps, template renderers and API platforms. Wireflow, rated best overall, links multiple image models on a node canvas and runs batch jobs through a single REST endpoint; OpenArt supports CSV/TXT batch generation of up to 500 images, the highest free batch limit; Recraft offers an asynchronous batch API at about $0.035 for raster images and about $0.08 for vector images.

  4. Basic data handling · ComfyUI创作项目发布22

    ComfyUI save nodes updated: preview support, pass-through output and workflow metadata embedding

    ComfyUI's "save IMAGE to file" and "save IMAGE+MASK to file" nodes are now compatible with the built-in "Save Image" node: images written to the output or temp folder are sent back to the frontend for preview and added to the generated images list, and a second `images` output has been added that passes through the original image.

Sep 18Fri
  1. ArchiGen AI · 图像生成与ComfyUI实战20

    AI Render Collaboration Canvas Review: Can Teams Track a Single Image's Revisions, Reviews, Approvals, and Archiving?

    AI rendering products are starting to emphasize shared canvases and team collaboration, but the evaluation focus should be the decision path rather than whether collaboration buttons exist. The article suggests trialing with five states—draft, shortlisted, checked, approved, and superseded—and testing unauthorized operations, permission revocation, and attribution records. Generation is not checking, comments are not approval, and downloading is not archiving; if a product cannot provide both creation history and approval history, teams need to separately decide where to store streamlined approval records.

  2. ArchiGen AI · 图像生成与ComfyUI实战22

    An Approved ComfyUI Render Is More Than a PNG: How to Preserve Workflows, Inputs, Environment, Checks, and Usage Decisions

    After a ComfyUI render is approved, the project record should include five types of files: the selected output, an explicit workflow JSON, an input manifest, an environment snapshot, and a review/usage record. The official example states that workflow information can be embedded in image metadata and dragged back into ComfyUI for restoration, but exporting, optimizing, or document processing may alter the file, so a file hash must be computed at approval time and written into the decision record.

  3. Theoretically Media65

    Theoretically Studio 4.0 Released: A Reusable AI Filmmaking Template

    Theoretically Media released Theoretically Studio 4.0, a reusable AI filmmaking template. The author used it to remake an old film made with Veo 2 in 2025, taking about a day and a half, using Seedance 2.5, GPT-6 Astra, and Runway's MCP connection during the process.

    Why it matters: The author remade an old film in a day and a half and publicly released a reusable AI production template and local dashboard, suitable as a reference for creators who want to build workflows.