@SSSS_CRYPTOMAN I noticed this as well and was going to start trying to figure out how to work around it but this is a great prompt to use. Thanks for sharing!
English
Ayo Shosanya
110 posts
Ayo Shosanya
@Shayo_95
Believer | Content Creator I love making things and sharing insights Founder of @sovereignLG & @himitsuministry
Katılım Mart 2026
75 Takip Edilen6 Takipçiler
ChatGPT Image 2.0
普通に作ると画像がギラギラになる問題をプロンプトで解決✨
このギラギラはおそらく学習段階で「キラキラしてる方がユーザーからの評価が高い」と判断された結果、エフェクトの過剰生成が起きてる現象だと思います
LoRAでCFGスケール高過ぎみたいな感覚でしょうか
確かにぱっと見は綺麗なんですが、画面全体に適用されると絵が散漫になり、かなり不自然です…
👇という事で対策プロンプトを作りました。これを最後に付け足すだけでOKです
---
グリッター・パーティクル・光の粒の散乱は多用せず、イラストとして必要な箇所のみに使う。細部を描き込みすぎて破綻しないよう注意して制作すること。
ｰｰｰ
画像1･3枚目は対策無しで生成、2･4枚目は対策プロンプト有りです
どうでしょうか？
元々の綺麗な光の描き方は残しつつ、見やすい絵に仕上がってると思います
このプロンプトは汎用的に使えると思うので是非やってみてください👍
ちなみに画風指示を「フラットシェーディング」とか「セルシェーディング」等にして改善もできますが、キャラの描き方がちょっと変わります
日本語
@hanagasa_ai What prompt did you use to generate the image of the girl? Curious about the art style. お願いします。
日本語
🍌プロンプトをChatGPT Images 2で。
細かい字までしっかり出るのはやはり堪らないものがありますねぇ。。
花笠@hanagasa_ai
みんなが作れば私がハッピーな気がしてきたので、ゲームの攻略本風のNano BananaProのプロンプト書いときますね 大したことは書いてないので気軽に作れますね！（Altに少し説明あり）
日本語
@OpenAI Any plans to expand this to normal consumers on Plus and Pro plans?
English
A hack many are already discovering with Codex!
Peter Gostev@petergostev
GPT Image 2 + Codex: or how to make Codex not suck at UI. Step 1: Generate a UI image (native in Codex) Step 2: Get Codex to implement the UI based on it Step 3: Get Codex to iterate until it aligns with the image as much as possible Codex is bad at initial UI, but very good at implementing a reference design, so this is your way out - iterate with the image model first and then Codex will do a good job.
English
Ayo Shosanya retweetledi
Voyage is the first AI-native RPG platform and signals the future of the $47b RPG market.
Built on our World Engine, nearly 60% of players call it revolutionary.
Backed by @NFX, @Griffin_GP, @AIFuturesFund, Midjourney, @AlbumVC + @Roblox execs.
@TechCrunch has the story 👇
English
Ayo Shosanya retweetledi
Arena Trends: Text-to-Image, Jan 2026 – Apr 2026
For most of the year, @GoogleDeepMind and @OpenAI traded the top spot within a tight margin - GPT-Image vs. Nano Banana - with the rest of the field clustered below 1,200.
Today, GPT-Image-2 breaks away with a score of 1,512, 242 points ahead of #2 Google.
The frontier continues to move.
Arena.ai@arena
Exciting news - GPT-Image-2 by @OpenAI has claimed the #1 spot across all Image Arena leaderboards! A clean sweep with a record-breaking +242 point lead in Text-to-Image - the largest gap we’ve seen to date. - #1 Text-to-Image (1512), +242 over #2 (Nano-banana-2 with web-search aka gemini-3.1-flash-image) - #1 Single-Image Edit (1513), +125 over #2 (Nano-banana-pro aka gemini-3-pro-image) - #1 Multi-Image Edit (1464), +90 over #2 (Nano-banana-2) No model has dominated Image Arena with margins this wide. Huge congratulations to @OpenAI on this major breakthrough in image generation! More performance breakdowns by category in the thread below.
English
Ayo Shosanya retweetledi
@CtrlAltDwayne That's their personal experience, not everyone has experienced that. I've spent no where near $200 for "max tier" and have been able to build my own tools and work on a great game idea. I've also never been stuck in week long loops. Not sure what the OP is doing tbh.
English
@haki_xer Yes it looks vibe coded. The purple color is a tell-tale sign. It also looks generic. I don't say that against you but that's just what the LLMs lean towards unless you steer them away from that. I've also dealt with that too.
English
Download this app at
screenkite.com
3x faster Screen Studio alternative.
With native AI Agent support
Editing your video using whatever AI you love
English
ScreenKite 1.0 Released. Welcome to the future.
Fastest macOS recording app with your favorite AI
Editing videos with Claude Code / Gemini / Codex
Auto transcription cut / b-roll with AI agents
Dynamic camera layout
video made with ScreenKite + HyperFrames
English
Ayo Shosanya retweetledi
Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude.
Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.
English
Ayo Shosanya retweetledi
@scaling01 It's 2x the cost of GPT 5.4 while edging over GPT 5.4 a little bit...all I'm going to say is if you thought the usage rate for Opus 4.6 was bad, just wait...
English
for all the people calling Opus 4.7 a mid update lmao
Vals AI@ValsAI
The new Opus 4.7 model places #1 on our Vibe Code Benchmark, at 71%. When we first released the benchmark 4.5 months ago, no model scored above 25%. This benchmark tests a model’s ability to create a fully functional web application from the ground up.
English
@vivoplt Hey! That looks exactly like my link for the app I built!
English
@LexnLin Not shocked at all 😂 With how much people have been complaining about the usage for 4.6, 4.7 isnt going to be any better.
English
I was about to buy a pro account to test Opus 4.7, but this is holding me back
Rand@rand_longevity
damn that happened fast
English