3 quick updates! We've updated the V7 model with improved image quality and coherence. There's a new lightbox editor that's easier to use. Last we've added a new experimental aesthetics parameter –exp (goes 0 to 100, 0 is default) that pumps up details and creativity. Have fun!
此次更新凸顯出 Midjourney 持續在 AI 圖像生成領域的創新腳步，展現其穩居產業領先地位的決心。
Exp 參數實際測試
提示詞(使用 V7 原始風格)
A young Taipei woman stands in a lush, green outdoor setting, surrounded by vibrant foliage and a rustic wooden structure in the background. The soft, natural light gently illuminates her face and attire, creating a warm and inviting atmosphere. She wears a delicate white blouse with lace detailing and ruffled sleeves, paired with black high-waisted pants adorned with gold buttons. Her dark brown hair is styled in loose waves, with bangs framing her face, adding to her youthful and cheerful appearance. She poses with her hands near her head, fingers slightly spread and relaxed, showcasing her well-manicured nails and graceful gestures. The focus is on her upper body and face, with a shallow depth of field that softly blurs the background, emphasizing her joyful expression and the intricate details of her outfit. The lighting is natural and diffused, casting soft shadows and highlighting the textures of her clothing and surroundings. --ar 16:9 --style raw --s 100 --p none
Google 於2025年4月29日宣布 AI 筆記應用 NotebookLM 的「Audio Overviews」功能現已支援超過50種語言，包括中文、法語、西班牙語、葡萄牙語、韓語、土耳其語和印地語等。此舉讓全球更多用戶能夠將筆記內容轉換為類似播客的音頻對話，提升學習與資訊吸收的效率。
NotebookLM 是 Google Labs 開發的 AI 筆記與研究助手，利用 Gemini 語言模型，協助用戶從上傳的文件中生成摘要、解釋和創意內容。「Audio Overviews」功能則能將這些內容轉換為由兩位 AI 主持人進行的音頻對話，模擬播客形式，讓用戶以更輕鬆的方式理解複雜資訊。
此次語言支援的擴展，使得來自不同語言背景的用戶也能享受這項功能。用戶只需在 NotebookLM 的設定中選擇「輸出語言」，即可生成對應語言的音頻摘要。此外，Google 也將「Audio Overviews」功能整合至 Gemini AI 聊天機器人和 Google Docs，進一步擴大其應用範圍。
Google 表示，這次更新是對「Audio Overviews」功能的早期探索，未來將根據用戶反饋持續改進，並計劃加入更多語言支援和功能強化。目前，該功能已在超過 200 個國家和地區推出，並逐步向更多用戶開放。
OpenAI 本週開始將 Deep Research Light 分批開放至 ChatGPT 免費帳戶，每位用戶每月可執行 5 次完整研究任務。此配額將於未來數週陸續到位，無須申請候補。Light 版採用 o4-mini 推論引擎與分層計算架構，官方目標是在降低約 60 % GPU 使用率的前提下，把產出時間控制在 2-5 分鐘。 Plus 與 Pro 層級維持現有高階配額，同時額外贈送 Light 版任務數，藉此提升免費用戶轉訂的機會。Google Gemini 與 Anthropic Claude 已經將類似研究代理功能下放至免費層，OpenAI 此次策略被業界視為「跟進同業的必要步驟」。OpenAI 表示，Light 版沿用完整版的 Source-Tracing Pipeline，所有輸出都會附上可點擊來源，並在檢測到潛在不實資訊時顯示黃標警示。
The lightweight version of deep research is powered by a version of OpenAI o4-mini and is nearly as intelligent as the deep research people already know and love, while being significantly cheaper to serve.
We're releasing an update to our image editor. It includes a refreshed UI, layers, a smart selection tool and smarter moderation. The editor works both via the lightbox on your MJ images or images you create from uploaded media. Last, all membership tiers now have access. Enjoy! pic.twitter.com/4a2fywa0yE
