Skip to content
Date and source

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

Oct 1Thu
  1. AImorph-艾摩弗形界 · 创作者专栏50

    Top-down typographic poster ceiling! 9 sets of AI object typographic posters, every character holds a memory

    A creator used Qwen-Image 3.0 to generate 9 sets of top-down still-life typographic posters, precisely assembling words such as "graduation," "going home," "keep going," "winter break," "teacher's grace," and "back to school" from real objects like graduation caps, train tickets, 2B pencils, and chalk, with each 3:4 image paired with a sticky note to highlight the theme. The same LoRA and image templates are now available on LiblibAI, and the prompts are public for reference and derivative creation.

Sep 30Wed
  1. 公众号:娱乐资本论22

    Bilibili danmaku AI judges, NetEase group-chat automatic enforcement: September internet patent roundup

    In September, major internet companies published patents密集ly: Bilibili applied for a livestream interaction patent that uses large models combined with danmaku to generate virtual judge commentary; NetEase applied for a group management patent where, after the group owner writes group rules, the system automatically parses them and issues real-time warnings and mutes. Youku is working on one-click script-to-storyboard splitting and video character expression editing, Baidu's ERNIE large model uses multi-agent automatic drafting and revision, Alibaba's Tongyi Wanxiang accelerates AI image generation, and Kuaishou's Kling diagnoses text-to-image shortcomings through layered analysis.

  2. note · catap_art3d · 图像视频创作分享34

    Trying Ming Image 0.1 Design in ComfyUI

    Ming Image 0.1 Design from inclusionAI under Alibaba-affiliated Ant Group has been added to the default templates of ComfyUI 0.38.0 stable. The template calls Qwen3.8 27B by default as a prompt enhancer; 16GB VRAM will OOM, so the author switched to running it with ollama and the comfyui-ollama node. The author considers its Japanese performance average, but its censorship is looser and its freedom higher than similar tools like Ideogram.

  3. ArchiGen AI · 图像生成与ComfyUI实战60

    ComfyUI Architectural Render Repair Order: Fix Walls First, Then Reflections

    ArchiGen AI proposes breaking "making it photorealistic" into six ComfyUI repair steps ordered by dependency: camera and subject geometry, material boundaries and scale, light direction and exposure, glass and reflections, plants and set dressing, and local defects and finishing. Each step requires saving numbered checkpoints and setting acceptance anchors.

  4. 阿真Irene · 创作者专栏78

    Alibaba Open-Sources Qwen-Image-2.1: One Model Covers Image Generation, Editing and Transparent Backgrounds

    Alibaba has open-sourced the Qwen-Image-2.1 image model, with a 7B visual generation component using a 32-layer Single-Stream DiT; one model covers generation, editing and native transparent-background images.

    Why it matters: The author presents extensive hands-on cases showing Qwen-Image-2.1's performance in image generation, editing and native transparency, and provides measured speeds on consumer-grade GPUs.

Sep 29Tue
  1. Adobe · 官方创作与影视博客58

    Adobe Adds Interactive Image and PDF Editing Controls to Its ChatGPT Plugin

    During OpenAI DevDay 2026, Adobe launched interactive editing features for its ChatGPT plugin, allowing users to select a portion of an image in the sidebar and have ChatGPT call the plugin to modify it, then fine-tune with dynamic sliders for hue, saturation, contrast, and more. Editing results are returned to the conversation as context.

  2. ArchiGen AI · 图像生成与ComfyUI实战26

    Don't Rush to Import a ComfyUI Workflow: Read This Diagram First

    When faced with a ComfyUI architecture workflow shared by a stranger, don't blindly install dependencies just to eliminate red node errors. Instead, start by tracing the output path backward from the Save Image node to understand inputs, constraints, variations, and dependencies. The article suggests filling out a six-line 'image passport' for each workflow, recording inputs, models, controls, variations, dependencies, and outputs, and distinguishing between structural controls and visual guidance, treating dimensions as a chain to verify node by node. The first test should be run on a non-confidential proxy image at a moderate resolution. Custom nodes must be treated as software installations, with repositories, versions, and licenses recorded.

  3. AI设计师Lena · 创作者专栏23

    10 Sets of SUMMER TIME Inflatable Balloon Letter Summer Poster AI Prompts

    A set of summer-themed poster prompts centered on 3D inflatable balloon-texture "SUMMER TIME" lettering, paired with real-life photos of athletic boys and girls in the sun with tennis, pools, and sunlight; the main colors are fluorescent green, peach pink, and sky blue. The lettering is puffy and three-dimensional with highlighted, reflective edges; backgrounds include tennis courts, pools, blue skies with white clouds, and palm trees; high saturation and high contrast, Y2K retro trend style, suitable for summer e-commerce posters, social media materials, and event banners.

Sep 28Mon
  1. note · tesalab · 图像视频创作分享22

    Runpod beginner generates an AI image while asking ChatGPT: 10 dollars topped up, 0.06 dollars spent on one image

    A low-spec PC user, with ChatGPT's guidance, completed a first AI image generation on the cloud GPU platform Runpod. He chose the Pods feature, the Runpod Stable Diffusion environment, and an RTX 4090 (0.74 dollars/hour); after topping up 10 dollars, he generated a 512×512 image of a red panda in a cafe, and the balance afterward was 9.94 dollars, for an actual cost of about 0.06 dollars.

  2. Midjourney-ai.hk · 中文提示词教程38

    Midjourney vaporwave prompt guide: pink-cyan neon and retro grids

    Midjourney vaporwave prompts work through four types of keywords together — pink-cyan gradients, CRT noise, Roman statues, and perspective grids — rather than just piling on neon pink purple. The article gives a four-layer prompt structure (subject, medium language, style anchors, parameters) and includes 4 copyable examples, all using --v 8.2 with --ar 3:4 or 16:9 and --stylize 170–250.

  3. ArchiGen AI · 图像生成与ComfyUI实战36

    Polishing rendered images requires an AI change map: reviewing ComfyUI image-to-image with a difference matte

    Addressing the problem that ComfyUI image-to-image polishing of rendered images also alters unauthorized areas, ArchGen AI proposes adding a difference matte to each review, comparing the aligned baseline image with the candidate image pixel by pixel, darkening unchanged areas and highlighting changed pixels. The workflow includes four steps: freezing the baseline image, defining the allowed-change matte, generating the raw difference, and separating allowed from unexpected changes, plus a flicker comparison; it classifies changes into four types — requested, necessary collateral, harmless incidental, and prohibited — rather than pursuing zero difference.

Sep 27Sun
  1. ArchiGen AI · 图像生成与ComfyUI实战26

    ArchiGen: unreviewed renders should not enter the AI refinement pipeline

    ArchiGen proposes setting six basic render acceptance gates before AI refinement: camera, geometry, materials, lighting logic, content, and evidence records; those that fail are sent back to the modeling software rather than handed to the model to patch. The article points out that ComfyUI is only a node interface, Flux is only a model family, and 4K only represents pixel dimensions—none of which guarantee architectural accuracy; once an image enters an image-to-image pipeline, existing defects become hard to attribute. Refinement is editing, not inspection, and retained effects must be measured against known inputs.

  2. AImorph-艾摩弗形界 · 创作者专栏48

    Stationery has an epiphany! 9 AI surreal posters turn pens and erasers into poetry

    A set of 9 stationery-themed AI surreal posters turns pens, erasers, and books into miniature scenes: a whale carrying a classroom, book pages bending into a glowing doorway, exam papers laid out as a marathon track. The images were generated with Qwen-Image 3.0; the same LoRA and image templates are now available on LiblibAI, and the full prompts are published alongside for reference and derivative work.

Sep 26Sat
  1. note · it_home_life · 图像视频创作分享58

    Hands-on with ChatGPT Images 2.5: capabilities and shortcomings after 100+ image sets

    The author tested ChatGPT Images 2.5 with over 100 prompt sets and believes the biggest improvement in realistic-style images is not realism but naturalness—less of the oily sheen and jarring composition typical of AI. Compared with a ramen image generated by Nano Banana, the latter has too much surface gloss, a stickier-looking egg yolk texture, and overall higher contrast. The author also notes shortcomings: insufficient variation in faces and hairstyles, fine mechanisms such as watch gears not meshing structurally, and easy breakdown in images with many elements such as crowds; the free version is usable, and the only difference from the paid version is the daily generation count.

  2. ArchiGen AI · 图像生成与ComfyUI实战38

    Stop generating logos from prompts: precise graphics need precise inputs

    ArchiGen argues that precise visual assets such as brand marks, wayfinding text and specified products should not be left to prompt generation but inserted into images as controlled inputs. The article uses Chaos's new Insert Images feature for Veras 5.0 as an example, explaining that users can place reference photos and logos on a 2D canvas and move, scale and rotate them, and the canvas layout is saved. The author recommends first building an asset package containing source files, versions, scope of use and placement notes, then locking the viewpoint, defining the plane, protecting assets and handling physical blending separately.

  3. ArchiGen AI · 图像生成与ComfyUI实战22

    AI relighting images are not daylighting analysis: ArchiGen proposes a five-sentence caption test

    ArchiGen AI points out that AI-relit interior images can present daylight effects with credible direction, softness and color temperature, but the model may not know the building's location, date, glass transmittance, surrounding obstructions and year-round weather data, so they cannot serve as evidence of daylighting performance.

  4. NerdyRodent · YouTube 创作视频75

    Qwen Image 2.1 running locally in ComfyUI

    Nerdy Rodent releases a tutorial on how to run Qwen Image 2.1 locally in ComfyUI. Besides text-to-image, the model also supports image editing with up to 10 reference images, for face swapping, head swapping, background swapping, multi-character composition and natural-language style changes; the INT8 model file is under 8GB.

    Why it matters: Demonstrates the text-to-image, image-to-image, inpainting and upscaling workflows of Qwen Image 2.1 in local ComfyUI, which can be used to evaluate a local image generation setup.

  5. AImorph-艾摩弗形界 · 创作者专栏36

    9 "Light in the Classroom" themed AI posters: generated with Qwen-Image 3.0, LoRA and template now available on LiblibAI

    A set of "Light in the Classroom" themed AI posters was generated with Qwen-Image 3.0, 9 in total, built around scenes such as late-night office lighting, a chalk whale leaping out of the blackboard, and chalk dust falling as snow, all in a 3:4 aspect ratio. The same LoRA and image template are now available on LiblibAI, and the full prompts have been published for reference and remixing.

Sep 25Fri
  1. Promptsref.com · 创作分享18

    This Week's Top Three Midjourney SREFs: Cinematic Lighting, Candy Colors, and Cartoon Attitude

    Promptsref's September 25 issue selected this week's top three Midjourney SREFs: cinematic anime hybrid, candy-colored 3D animation hybrid, and graphic cartoon single-code style. All three center on character design, suited respectively to visual novel scenes, mascot concepts, and avatar design. The official note clarifies these are only creative direction references, not actual outputs.

  2. Leeveo AI Tutorials · 创作教程与项目分享18

    ChatGPT /command 001: Use /leeveo to Generate Multi-Style Portraits While Maintaining Character Consistency

    ChatGPT's /leeveo command can transform the same character into different faces, ages, outfits, expressions, poses, and story worlds while maintaining identity consistency. The official demo uses Gandalf, Frodo, Mad Max, and The Bride as references, with one reference character branching into ten creative directions for character building in AI films, storyboards, games, marketing campaigns, and visual narratives.

  3. ArchiGen AI · 图像生成与ComfyUI实战24

    Why ComfyUI Architectural Rendering Tutorials Should Show Failure Cases: A Reproducible Teaching Framework

    In response to community demand for ComfyUI architectural rendering tutorials, ArchiGen proposes that tutorials should show failure frames rather than only finished results: each failure frame isolates one erroneous behavior, records the corresponding settings, and provides the next correction. The tutorial recommends starting from the official image-to-image baseline, adding only one architecture-related control at a time and queuing for verification, first teaching four types of failures—geometry loss, control conflicts, insufficient changes, and local repair—before producing the final render.

Sep 24Thu
  1. ArchiGen AI · 图像生成与ComfyUI实战32

    ComfyUI Rendering Debugging Guide: Verify Installation First, Then Check Workflow, Conditional Control, and the Architecture Itself

    For architects who want to use ComfyUI to improve rendering without losing composition, ArchiGen proposes troubleshooting in four sequential layers: installation, workflow, conditional control, and architecture. An error in an earlier layer invalidates conclusions from later layers. The installation layer requires confirming that the model and VAE are loaded and that the same settings can run successfully twice in a row. The conditional control layer recommends fixing all other parameters and only adjusting denoise, then testing control strength separately, to avoid adjusting two variables at once.

Sep 23Wed
  1. note · welina_kokua · 图像视频创作分享12

    No Experience Needed: Copy-and-Paste AI Image Generation Prompts to Instantly Elevate Your Blog and SNS

    For blog and SNS creators, a collection of ready-to-copy English AI image generation prompts has been compiled. Paste them into ChatGPT (DALL-E 3) or Midjourney to generate images. The prompts cover five scene types: coffee table, watercolor woman, flat business illustration, pastel plant background, and Nordic desk. Three rules are provided: input in English, specify a style, and specify an aspect ratio of 16:9 or 1:1.

  2. note · ready_whale7197 · 图像视频创作分享8

    Can AI-Generated Images Serve as Work Deliverables? Reflections After a Hands-On Session

    The author tried taking on freelance work using AI-generated images and found that "being able to generate" and "being able to deliver as work" are two different things: taking on orders also requires understanding client needs, adjusting dimensions, arranging text and layout, handling revisions, and delivering a finished product that is easy for the client to use. AI only handles image generation; deciding "what to do" still relies on humans. The author believes AI image creation has potential for challenges, but is not yet sure whether it has reached a level suitable for taking on orders.

  3. note · genkidesu2021 · 图像视频创作分享62

    Using 38 English style prompts to eliminate the AI look from image generation

    The author compiled 38 English style prompts that can be appended after Japanese drawing descriptions to lock in an art style, covering categories such as hand-drawn, watercolor, flat, printmaking, and manga. The article also provides four techniques for removing the AI look, including explicitly requesting slightly irregular, adding physical texture parameters such as paper grain, using limited 3 to 5 color palette to restrict color usage, and setting AVOID negative items to prohibit plastic-looking effects and excessive gradients, along with a copy-ready universal template.

  4. Adobe · 官方创作与影视博客62

    Adobe completes acquisition of Topaz Labs, integrating AI image quality enhancement into Firefly and Photoshop

    Adobe announced the completion of its acquisition of AI video and image enhancement company Topaz Labs, which will retain its independent brand, with its applications and models continuing to be offered as standalone products. Topaz Labs' technology has been integrated into Adobe Firefly and Photoshop for super-resolution and enhancement of images and videos, with its Neurostream technology supporting processing of complex AI models on local devices or Topaz's cloud.

    Why it matters: Adobe integrates Topaz Labs' image quality enhancement models into Firefly and Photoshop, offering insight into the integration trends of post-production toolchains.

  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.

Sep 22Tue
  1. Oimi · 中文图像视频教程62

    10 AI Portrait-Grade Prompts (2026)

    Oimi publishes 10 AI portrait prompts, each accompanied by the original English text, a Chinese translation, and a real generated example image. The styles covered include golden hour rooftop, geometric window light, studio color clash, rainy night café window, Rembrandt black and white, 90s direct flash, long-exposure street, and oil painting.

  2. 腾讯混元 · 公开公众号 RSS72

    Tencent Hunyuan Releases Hy Image3.5 Preview, a Professional-Grade Image Generation Model

    Tencent Hunyuan releases Hunyuan Image 3.5 Preview (Hy Image3.5 preview), positioned as a high-value professional-grade image generation model. It supports text-to-image and image-to-image generation, up to 5 reference images per upload, up to 2K resolution output, and multi-turn conversational creation.

    Why it matters: Tencent Hunyuan releases a preview of its professional-grade image generation model, providing 2K output, API pricing of 0.15 yuan per image, and internal blind test comparison results, which can be used to assess cost and image generation capability.