Skip to content
Date and source

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

Oct 22Thu
  1. note · kawakijourney_ai · 图像视频创作分享30

    Reproducing the process of an office lady going to a hotel after work to change into Halloween cosplay using Midjourney/ChatGPT/Gemini (with prompts)

    A Japanese creator has released a set of prompts for reproducing the full process of "an office lady walking into a hotel after work and changing into Halloween cosplay" using Midjourney, ChatGPT, Gemini, and other AI generation tools, broken down from STEP1 to STEP5: walking out of the office building wearing an employee badge, walking on a night street, taking off shoes in the hotel entrance, and cosplay poses on the sofa and bed.

  2. Frank Houbre · AI电影制作实战38

    Best AI director software ranking for 2026: comparison of six tools including ScreenWeaver, LTX Studio, and Saga

    In the 2026 AI director software ranking, ScreenWeaver ranks first, followed by LTX Studio and Saga. The ranking evaluates six tools covering at least two production stages based on criteria such as "whether generated shots remain linked to the script text." ScreenWeaver integrates 7 video models including Seedance 2.5, Veo 3.1, and Kling 3.0; the Auteur version costs 39.99 USD/month, but the tool does not include full editing.

  3. Ciaro · 中文AI电影制作教程38

    Google AI Studio video generation test: evaluate with continuous shots rather than highlight clips

    When testing video generation with Google AI Studio, you should produce a three-shot continuous sequence with editing decisions rather than looking only at a single highlight clip. The test needs a unified shooting brief, a limit on attempts per shot (e.g., three) and a preset cost cap, and should check whether character identity, prop positions and spatial direction remain consistent across the edit points. Review revisions should regenerate only the affected shots, and the adoption criterion should be delivering an editable sequence at an acceptable cost.

  4. Luma News16

    Gemini Omni Flash vs. Wan 3.0: Which Is Currently the Best AI Video Model?

    Gemini Omni Flash excels at conversational editing and multimodal input. Wan 3.0 can generate clips up to 30 seconds long and supports document input. Both have their own advantages. Luma AI's Ray 3.2 offers multi-keyframe sequences, HDR output, and EXR export. Layers (powered by Uni-1) supports single-element modification, making it suitable for production workflows. In 2025, the AI video generation market size is approximately $700 million to $2 billion, with North America leading.

  5. Social Media Today45

    YouTube releases report: how creators use AI tools

    YouTube commissioned NRG to survey nearly 700 users aged 14-44 who uploaded videos in the past 12 months; 72% of surveyed uploaders said they use AI on videos or images. More than two-thirds of AI video publishers said they themselves appear in the content, and 61% of respondents are interested in using AI to insert themselves into video footage. The report also mentions creators using generative AI to build complete film worlds, forming a group of "AI authors."

Oct 21Wed
  1. Higgsfield 官方博客38

    A hands-on comparison of the top ten AI video generators in 2026: Higgsfield, Runway, Kling, and more

    Higgsfield ranks first in the 2026 comparison of the top ten AI video generators, integrating 35+ models and tools for character consistency, ad production, and agent capabilities into one subscription. Runway focuses on editing and VFX workflows with Gen-4.5, the official Kling app offers direct access to Kling 3.0/4.0, and Google Flow offers Veo 3.1 starting at $4.99. All platform prices and quotas were verified in October 2026.

  2. Krea 官方博客62

    Krea 3 vs Nano Banana Pro: leads on leaderboards, but still behind on 4K and long text

    Krea officially compared Krea 3 with Nano Banana Pro, stating that Krea 3 ranks higher on all three leaderboards—Arena Multi-Image Edit, Design Arena and Artificial Analysis—that most edits return within 8 seconds at native 2K, and that it is 3 times cheaper than GPT Image 2 and Nano Banana Pro.

    Why it matters: Krea officially compared its own Krea 3 with Nano Banana Pro using three public leaderboards and same-prompt tests, and listed the applicable scenarios for each.

  3. Krea 官方博客60

    Krea 3 appears on three image-editing leaderboards: a breakdown of each ranking and interval

    Krea 3 Ultra ranked 4th in Arena multi-image editing, 8th in single-image editing, 5th in Artificial Analysis AA-Image-Editing v2.1, and 4th in Design Arena (out of 51 models) on three image-editing leaderboards published on October 19-20.

    Why it matters: Krea officially breaks down the votes, confidence intervals, and ranking intervals of the three leaderboards item by item, explaining why the same model appears in four positions.

  4. 宝玉的分享 · 创作与项目分享42

    NotebookLM Video Overview system prompt: the complete creation workflow

    NotebookLM's Video Overview creation skill file has been made public, fully presenting a multi-step workflow from reading notebook sources, helping the user select a topic, and confirming settings to submitting to the video generation pipeline. The prompt focuses on intent disambiguation: distinguishing creating a new video from asking about an existing video, discarding old options when the user restarts, and submitting the user-confirmed topic text verbatim for generation. The file also includes schemas for questionnaires and review cards, formatting and visual style rules, localization requirements, and example tool calls.

Oct 20Tue
  1. Frank Houbre · AI电影制作实战67

    How to keep the same character consistent across shots in AI video

    Frank Houbre gives a production method for keeping the same character across shots: first make a front-facing main image plus two or three angle references, then use the model's built-in character reference entry point rather than pure text prompts, and finally break down the storyboard by shot risk.

    Why it matters: The author horizontally compares the character reference entry points of Kling, Seedance, Flow, and Runway according to official documentation, and provides a reusable storyboard and reference image preparation process.

  2. note · aiholic0831 · 图像视频创作分享40

    How big is the difference between 1K and 2K resolution in image generation AI? Upscaling comparison with Grok using the same prompt

    Using the Grok API's grok-imagine-image-quality model at a 3:4 ratio with the same prompt to generate 1K (864×1152) and 2K (1776×2368) images for comparison, it was found that 1K is already sufficient for close-up faces, while the difference is obvious in compositions where the person is smaller, such as full-body shots: at 1K eyelashes blur into line blocks and pearl earrings become blurry dots, while 2K can draw eyelashes individually and retain highlights.

  3. Wireflow · AI短剧与视频制作教程65

    Field test of 8 Veo 3 prompts on Veo 3.1: 5 passed, 1 partially passed, 2 failed

    Wireflow ran each of 8 Veo 3 prompts once with Veo 3.1 Fast, at 8 seconds, 720p, with sound; the results were 5 passes, 1 partial pass, and 2 failures. The passes were mostly single shots with a single action plus one line of dialogue under 15 words; the failures centered on requiring three shot changes within one 8-second clip, and a 39-word line being truncated to 31 words.

    Why it matters: The author tested 8 Veo 3.1 Fast prompts with the same set of parameters, giving the reason for success or failure and a rewrite for each, which can be compared against your own video production workflow.

  4. 我爱计算机视觉 · 公开公众号 RSS50

    WorldCast tests 12 visual generation models: can they draw a chair, and can they also "fake" a chair?

    WorldCast built a benchmark covering 8 task categories, 320 handcrafted cases, and 1,352 visual states to be generated, evaluating 12 image models, unified multimodal models, and video generation systems. GPT-Image-2.5 paired with GPT-6 Astra scored 72.8, surpassing the participating video generation systems; when switched to outputting all steps at once as a grid image, the score dropped to 50.1.

Oct 19Mon
Oct 18Sun
  1. note · zou_aidouga · 图像视频创作分享42

    25 AI video camera movement prompts from Google's official guide: 13 camera angles + 12 camera movements, with a ChatGPT-generated comparison cheat sheet

    Based on the 13 Camera angles and 12 Camera movements in Google Cloud's video generation prompt guide (updated October 15), the author used ChatGPT to generate comparison images of the same woman, the same alley, and the same sunset, making a cheat sheet of 25 camera movement names.

  2. note · silver517 · 图像视频创作分享18

    Why AI drawing blocks "A Little Princess Sara": SakanaAI explains search term contamination

    In AI drawing, both "小公女セーラ" and its English name "A Little Princess" are blocked. SakanaAI explains that the name has been contaminated on the internet: some people used AI to generate malicious images of the character and used them on malicious click sites. When the author tried to view the truth site provided by SakanaAI, the browser displayed a phishing scam warning.

Oct 16Fri
  1. note · singtothesky79 · 图像视频创作分享15

    Self-taught AI video making ④: the unexpected unfolding from writing a novel with Gemini to manga, animation and a theme song

    A self-taught AI video creator used Gemini to generate a Japanese sword-themed novel in 6 parts, then tried to adapt it into manga and animation. The manga stage yielded only storyboards and images that couldn't form a complete work, but the image quality was very high; after switching to animation, they succeeded in making a "moving manga," and after adding color and voiceover, a theme song was unexpectedly generated automatically.

  2. 极客公园66

    Character.AI files for bankruptcy reorganization; Disney bids for its IP licenses with a $20 million floor price

    On October 12 local time, Character.AI filed a Chapter 11 reorganization petition with the U.S. Bankruptcy Court for the District of Delaware, initiating a court-supervised asset auction. Disney, as the stalking-horse bidder, offered $20 million to acquire a non-exclusive perpetual license for part of the intellectual property, and committed to providing transitional technical services using its former employees. The transaction does not include the company entity, the product app, or user assets.

    Same story, the pick is 《Character.AI files for bankruptcy, Disney plans to bid $20 million for its assets》

  3. Wireflow · AI短剧与视频制作教程48

    How Claude generates video via MCP: a pipeline from script to finished film

    Claude itself cannot generate video, but it can call video models via MCP and return a finished MP4. In a four-node pipeline built on Wireflow, Nano Banana Lite generates keyframes, Gemini Omni Flash 1.1 and Kling O3 each produce 5-second clips, and Video Editor composites a 1920x1080 finished film. The whole process takes 101 seconds and consumes 176 credits.

  4. VibeClips · AI视频创作教程20

    Privacy and security of AI video generators: how VIBE handles your photos, prompts and videos

    VIBE's privacy policy stipulates that uploaded photos are only sent to the AI model APIs needed to generate videos (such as Kling, Veo, Sora, Seedance, WAN), are not used to train models, are not used for facial recognition, are automatically deleted after being stored for a maximum of 90 days, and users can also manually delete them at any time. Prompts and generated videos are retained in the account until the user deletes them, and personal data is not sold; the terms of service grant users a license to use generated videos for personal and commercial purposes, and users retain ownership of the original prompts and images.

Oct 15Thu
  1. The Verge · AI48

    Celebrity AI face-swapping proliferates: Instagram blue-check account posts fake Mia Goth images, Meta only partially cleans up

    An Instagram account with over a million followers and blue-check verification posted AI-generated fake images of Mia Goth that received nearly 500,000 likes; after spreading through gossip media and Google search AI summaries, Instagram deleted the account on the grounds of impersonating a public figure.

  2. note · bold_oxalis9933 · 图像视频创作分享10

    Wanting the cat to strike a JOJO pose but running out of AI image generation quota: lessons from Canva, Google AI Mode and English prompts

    A user tried to use AI image generation to make their cat strike a JOJO pose, repeatedly adjusting Japanese prompts without success, used up their free quota, and ultimately failed to complete it. During the experience, Canva's generated image quality was excellent, but free services are generally limited by quotas; after switching to Google Search's AI mode, they felt its image generation restrictions were more lenient than Gemini's.

  3. note · tidaa · 图像视频创作分享42

    How to use Gemini Omni: Generate a CM video with 5 images and a storyboard, then edit it into a finished piece with Canva (Part 2: Video Generation and Editing)

    Using 5 key visual images and a storyboard generated with Midjourney, the author generated 5 vertical 9:16 video clips in Gemini Omni, each approximately 3 seconds long, corresponding to five shots: hook, opening the bottle, drinking, whip pan to the sea, and beach afterglow. The author then assembled the 5 video clips into a single CM in Canva, originally planned for 15 seconds, with a final cut of 20 seconds.

  4. note · apt_zebra8914 · 图像视频创作分享22

    ChatGPT vs. Gemini image generation comparison: 5 rounds of testing with the same prompts, Gemini scores 70 vs. ChatGPT's 65

    Under the same set of prompts, ChatGPT (Pro, thinking level "high") and Gemini (Nano Banana 2.1) scored 65 and 70 respectively across five tests: thumbnails, photorealistic style, funny comics, posters, and partial edits. ChatGPT tends to "polish results to look better," deviating from the specified style in photorealistic and comic categories; Gemini is more faithful to the prompts, but the emphasis treatment of the word "そう" in the poster was penalized. Both scored full marks on partial edits.