agentic-media Trend Report
Dashboard へ戻る
3
High impact
5
Mid impact
52
Signal watch
なぜこのサイトを作ったのか
私たちそれぞれが個別にAIを使って情報収集し、同じような3行要約を作るたびに、世界中で膨大な電力と計算リソースが消費されています。 本プロジェクトは、あらかじめ広範な情報を取得・集約しておくことで、個別のAI実行回数を減らし、地球環境（GPU/TPU負荷）に配慮した効率的な情報収集を目指す実験的なダッシュボードです。
Recent Commits to hyperframes:main
feat(catalog): put the variables panel back, on payloads (#3199)
・feat(catalog): put the variables panel back, on payloads (#3199) * feat(catalog): put the variables panel back, on payloads The panel drove its preview by loading an .html from docs/public, a type the host does not publish, so it showed an empty frame in production and was parked when the catalog was re-landed.
・It now mounts the same JSON payload the plain player uses and re-mounts it as values change, injecting them
Recent Commits to hyperframes:main
feat(studio): ask for feedback when a render ends, not on a session c…
・feat(studio): ask for feedback when a render ends, not on a session counter (#3205) ## What Studio's feedback prompt now fires when a render finishes or fails, instead of on a session counter, and the reports it collects carry enough context to act on.
・- Replaces the 32px inline bar with a card in the existing toast stack - Adds `studio_feedback_shown` / `studio_feedback_dismissed` / `studio_feedback_interview_click`
#動画生成タグ
MiniMax H3とは？音声同時生成の2K動画AIを世界一わかりやすく解説【2026年8月最新】
・はじめに：また新しい動画生成AIが出た、で終わらせない 2026年7月31日、中国MiniMaxが「MiniMax H3（通称：Hailuo 3.0）」を発表しました。さらに8月3日には、モデルの重み（ウェイト）がHugging Faceで公開され、動画生成AI業界では珍しい「オープンウェイト」の選択をしています。
・Sora、Veo、Klingと動画生成AIが乱立するなか、なぜH3がいま注目されているのか。この記事では、H3の技術的な特徴・料金・商用利用のライセンス条件・競合モデルとの違いまで、2026年8月時点の最新情報をもとに整理します。
・MiniMax H3の3つの特徴 1.
Recent Commits to remotion:main
`@remotion/player`: Add `_experimentalKeepAudioContextAlive` prop (#1…
・`@remotion/player`: Add `_experimentalKeepAudioContextAlive` prop (#10313) * Add keepAudioContextAlive Player prop An opt-in for editor-style applications (NLEs, DAW-like tools) that play and pause many times per session.
・When enabled, the shared AudioContext is started once and never suspended: - suspend() silences through the gain node instead of suspending, so the context clock keeps running and already-scheduled
Recent Commits to hyperframes:main
feat(studio): show every colour of a mixed selection in the swatch (#…
・feat(studio): show every colour of a mixed selection in the swatch (#3144) * feat(studio): show every colour of a mixed selection in the text swatch Selecting text painted in more than one colour showed a white swatch.
・The toolbar reads a property only when the whole selection agrees on it, which is right for bold and italic (a toggle is on or off) but wrong for a swatch: with nothing to report it fell back to the de
Recent Commits to hyperframes:main
fix(producer): mix audio into a container that can record encoder del…
・fix(producer): mix audio into a container that can record encoder delay (#3200) * fix(producer): mix audio into a container that can record encoder delay Every rendered composition's audio landed 1024 samples (21.33 ms at 48 kHz) after its authored `data-start`, against a frame-accurate video track.
・The mix is AAC-encoded, and AAC encoders emit ~1024 priming samples.
・The mix was written to a raw ADTS `.aac` file, whi
#動画生成タグ
【MiniMax H3】RTX 3060(VRAM 12GB)で試した――2人別々のダンス、別音声リップシンク、専用アプリ化まで
・MiniMax H3は、文章だけでなく、画像・動画・音声を組み合わせて参照できる動画生成AIです。公式にも、異なる素材へ「キャラクター」「動き」「カメラ」「声」といった役割を割り当てられるモデルとして紹介されています。
・今回試したかったのは、単純な画像からの動画生成ではありません。
・・2人のキャラクターへ、それぞれ別のダンスをさせる ・キャラクターの動きとは別に、用意した音声でリップシンクさせる ・複雑なComfyUI操作を、専用アプリから簡単に実行する ・RTX 3060 12GBでも、できるだけ現実的な時間で生成する このあたりを目標に、試行錯誤してきました。
Release notes from remotion
v4.0.508
・What's Changed @remotion/studio: Add 3D transform controls by @JonnyBurger in #10278 @remotion/studio: Add "Open in Terminal" option by @JonnyBurger in #10309 @remotion/studio: Add Finder to "Open in" menu by @JonnyBurger in #10231 @remotion/studio: Show Chrome Headless Shell download progress by @henriquepe in #10268 @remotion/studio: Show zod .describe() text as tooltip in the props editor by @henriquepe in #10254
Recent Commits to remotion:main
`@remotion/mac-cursors`: Support interactive cursor keyframes (#10330)
・`@remotion/mac-cursors`: Support interactive cursor keyframes (#10330) * Refactor canvas capture preview cursor * Support hold keyframes for cursor enums * Inline interactive macOS cursors * Hardcode canvas capture cursor keyframes * Make macOS cursors draggable in Studio * Use style scale for macOS cursors * Apply capture density to canvas cursor * Keyframe canvas cursor press scale * Keyframe canvas cursor position
Recent Commits to remotion:main
`@remotion/remotion-media`: Make catalog crawlable (#10322)
`@remotion/remotion-media`: Make catalog crawlable (#10322)
Recent Commits to remotion:main
`@remotion/remotion-media`: optimize content negotiation for coding a…
・`@remotion/remotion-media`: optimize content negotiation for coding agents fetch calls (#10307) * `@remotion/remotion-media`: Negotiate Markdown catalog * `@remotion/remotion-media`: Document production negotiation
Recent Commits to remotion:main
`@remotion/studio`: Add "Open in Terminal" option (#10309)
・`@remotion/studio`: Add "Open in Terminal" option (#10309) * `@remotion/studio`: Add default terminal setting * `@remotion/studio`: Use native Ghostty artwork * Restore brand Remotion config * List all installed terminals in Open in menu * Remove default terminal setting and fix Ghostty launch * Make terminal menu inputs explicitly nullable * Fix Warp terminal working directory
Recent Commits to remotion:main
`@remotion/studio`: Add 3D transform controls (#10278)
`@remotion/studio`: Add 3D transform controls (#10278)
Recent Commits to remotion:main
`@remotion/studio`: Fix menu Escape race (#10317)
`@remotion/studio`: Fix menu Escape race (#10317)
Recent Commits to remotion:main
`@remotion/studio`: Preserve 3D rotation cursor while dragging (#10321)
`@remotion/studio`: Preserve 3D rotation cursor while dragging (#10321)
Recent Commits to remotion:main
`@remotion/studio`: Respect disconnected state in composition control…
・`@remotion/studio`: Respect disconnected state in composition controls (#10315)
Recent Commits to remotion:main
`remotion-monorepo`: Fix pre-commit hook on Linux (#10320)
・`remotion-monorepo`: Fix pre-commit hook on Linux (#10320) Co-authored-by: anthod0 <91065936+ocherry341@users.noreply.github.com>
Recent Commits to remotion:main
`remotion`: Re-check audio resume success before reporting failure (#…
・`remotion`: Re-check audio resume success before reporting failure (#10311) * `remotion`: re-check audio resume success before reporting failure waitUntilActuallyResumed() polls for output-clock progress with requestAnimationFrame but declares failure with a wall-clock setTimeout.
・When the main thread is busy for about a second (heavy rendering, a bundler rebuild in development), the timeout fires before the rAF call
Recent Commits to hyperframes:main
chore: regenerate the skills manifest (#3203)
・chore: regenerate the skills manifest (#3203) The hyperframes-cli skill gained the search-miss report and its content hash moved with it.
・The manifest is the published freshness fingerprint, so it has to be regenerated in the same change.
Recent Commits to remotion:main
Elements: Refine Name Lower Third (#10324)
・Elements: Refine Name Lower Third (#10324) * Refine Name Lower Third element * Publish Name Lower Third previews
Recent Commits to hyperframes:main
feat(catalog): group the sidebar by what you came to make (#3194)
・feat(catalog): group the sidebar by what you came to make (#3194) * feat(catalog): group the sidebar by what you came to make Nineteen shelves in one column is a list to read, not a menu to scan, and two of them were oversized: Effects held 54 items and Showcases 38, two thirds of which were near-identical code-snippet themes.
・The shelves now sit under eight openable sections named for what a reader is making, and th
Recent Commits to hyperframes:main
feat(cli): keep your edits when you reinstall a catalog item (#3193)
・feat(cli): keep your edits when you reinstall a catalog item (#3193) * feat(cli): keep your edits when you reinstall a catalog item Running add again overwrote whatever was on disk, so a project that had tuned an installed block lost that work without being asked or told.
・The installer now records a hash of each file as it installs it, and compares before writing.
・A file that still matches is replaced as before; one
Recent Commits to hyperframes:main
feat(core): sanitize rich text on the way into a composition (#3141)
・feat(core): sanitize rich text on the way into a composition (#3141) * feat(core): sanitize rich text on the way into a composition Studio's patch vocabulary was inline-style, attribute, html-attribute and text-content.
・text-content assigns textContent, and the text-field model escapes markup on the way out and refuses a change in child structure, so a styled span had no route into a composition file.
・Adds a rich-tex
Recent Commits to hyperframes:main
feat(registry): bring back the video-primitive moves (#3169)
・feat(registry): bring back the video-primitive moves (#3169) Restores the 208 catalog items reverted after their previews 404'd in production, this time on the payload mechanism rather than the .html files that caused the outage.
・The generator no longer writes a preview document to docs/public.
・That writer, and the machinery under it, existed only to produce files the docs host discards, so it is gone rather than byp
Recent Commits to hyperframes:main
feat(studio): apply a style to a run of characters (#3142)
・feat(studio): apply a style to a run of characters (#3142) * feat(studio): apply a style to a run of characters Styling text in a composition cannot be done by wrapping a DOM range in a span.
・That is three lines, and then every interesting case is a special case: recolouring nests spans that shadow each other, removing a style cannot reach the ancestor that set it, and styling across an existing run's boundary has to
Recent Commits to hyperframes:main
feat(studio): edit and style text in the preview (#3143)
・feat(studio): edit and style text in the preview (#3143) * feat(studio): edit and style text in the preview Double-press a text element in the canvas and the caret opens where you pressed, in the element itself rather than in a panel.
・Select characters and a small toolbar offers colour, bold, italic and underline, applied to exactly those characters.
・The toolbar lives in Studio's document rather than the composition'
Recent Commits to hyperframes:main
fix(catalog): keep the recorded video for canvas drawElement previews…
・fix(catalog): keep the recorded video for canvas drawElement previews (#3171) Eight blocks paint DOM into a canvas with ctx.drawElementImage(), which sits behind chrome://flags/#canvas-draw-element.
・A reader without the flag gets a preview that mounts, plays, and shows an empty canvas, which is what the 3D device blocks were doing even after their models were embedded correctly.
・The recorded video was captured by a r
Recent Commits to hyperframes:main
fix(catalog): resolve assets a script loads by name (#3170)
・fix(catalog): resolve assets a script loads by name (#3170) The 3D blocks fetch their model with loader.load("models/iphone.glb"), which no attribute or url() pattern can see, so the path survived into the payload untouched.
・Inside a srcdoc iframe it resolved against the docs page and 404'd, leaving vfx-iphone-device and ios26-liquid-glass mounted and playing with an empty scene.
・Names that resolve to a real file in
Recent Commits to hyperframes:main
fix(cli): route HeyGen API calls through canary (#3201)
fix(cli): route HeyGen API calls through canary (#3201)
Recent Commits to hyperframes:main
fix(core): key the preview volume envelope to the clip, not the timel…
・fix(core): key the preview volume envelope to the clip, not the timeline (#3198) A GSAP volume fade on an audio clip that starts after t=0 left the preview silent for the clip's whole length while the encoded render was correct.
・`mediaVolumeEnvelope` is meant to keep preview and render on one envelope, and its contract is "normalise, then read with track-relative seconds".
・The preview skipped both halves.
Recent Commits to hyperframes:main
fix(core): stop a graded plate painting through an inactive clip (#3196)
・fix(core): stop a graded plate painting through an inactive clip (#3196) A color-graded image or video inside a timed sub-composition kept painting after its clip window closed.
・The runtime hid the sub-composition wrapper with `visibility: hidden`, but the grading canvas carried an explicit inline `visibility: visible`, and an explicit value on a descendant escapes an ancestor's inherited `hidden`.
・The treated plate
Recent Commits to hyperframes:main
fix(studio): match the write receipt in dev, so an edit stops reloadi…
・fix(studio): match the write receipt in dev, so an edit stops reloading the preview (#3206) ## What Editing anything in the canvas on the dev server reloaded the preview iframe.
・It no longer does.
・## Why The write receipt exists to prevent exactly this: Studio marks its own writes so the file-watcher echo can be told apart from somebody editing the file underneath it.
Recent Commits to remotion:main
Internal tests: Remove flaky Tailwind v4 watch test (#10316)
Internal tests: Remove flaky Tailwind v4 watch test (#10316)
Recent Commits to remotion:main
Move news article highlight to Storytelling (#10326)
Move news article highlight to Storytelling (#10326)
Recent Commits to remotion:main
Pause playback when trimming (#10312)
Pause playback when trimming (#10312)
Recent Commits to remotion:main
Preserve inspector scroll when adding effects (#10328)
Preserve inspector scroll when adding effects (#10328)
Recent Commits to remotion:main
remotion.dev/convert: Scale cursor while pressed (#10329)
remotion.dev/convert: Scale cursor while pressed (#10329)
Recent Commits to hyperframes:main
revert(cli): keep HeyGen API traffic on stable prod (#3202)
・revert(cli): keep HeyGen API traffic on stable prod (#3202) * Revert "fix(cli): route HeyGen API calls through canary (#3201)" This reverts commit 5545521556c232435e959e3fd9b6ab1220278ea3.
・* fix(cli): remove remaining EF canary routes
#音声合成タグ
「365日の紙飛行機」（AKB48）ボカロ初音ミク・カバー NHK連続テレビ小説『あさが来た』主題歌
「365日の紙飛行機」（AKB48）ボカロ初音ミク・カバー NHK連続テレビ小説『あさが来た』主題歌
#動画生成タグ
【1.6万文字で徹底解説】 実作業10〜15分でバズるAI動画を作る方法を、プロンプトごと全公開します
・※ 期間限定で特別価格1480円で販売 このアカウントで投稿しているAI動画を見て、 続きをみる
#動画生成タグ
【2026年版】MiniMax H3をComfyUIで始める方法①｜初心者向けにインストールから徹底解説
【2026年版】MiniMax H3をComfyUIで始める方法①｜初心者向けにインストールから徹底解説
#動画生成タグ
⚔️反論させたうえで自分の出した回答を何度も修正させていく自己批評ループの具体的な組み方🌀導入前後で回答の精度がどこまで変わったかを実例つきで検証📈
・🔥その「一発回答」、実はまだ本気を出していません ChatGPTに質問して、返ってきた答えをそのままコピーして使う。
#動画生成タグ
🌏Fable5とMythos5が6月にいったんアクセス停止となり、7月1日に再開するまでの一連の経緯、⚖️公式の発表をもとに一次情報から丁寧に追いかけた記録をまとめる🗞️
・🚨最先端AIが、ある日突然「使えなくなった」 —— その裏側を知らないままでいいですか？ 2026年6月、AI業界に静かな衝撃が走りました。
#動画生成タグ
🤖TikTok投稿の台本をChatGPTに書かせて失敗した経験から見えてきた、視聴者に刺さる冒頭3秒を作るための具体的な指示文の作り方を、実例つきで公開していきます📝
・🔍「なんとなく良い感じの台本」ではもう視聴者は止まってくれない、その理由とは？ 「ChatGPTにTikTokの台本を書いてもらったのに、なぜか再生数が伸びない」 続きをみる
#音声合成タグ
40年前のCokeのノベルティをmicro:bitラジコンに魔改造した話の続き。音声合成でしゃべれるようにした
40年前のCokeのノベルティをmicro:bitラジコンに魔改造した話の続き。音声合成でしゃべれるようにした
Zennの「音声合成」のフィード
TTS の分割位置は実時間比から決まる（測定環境を汚したまま設計して痛い目を見た話）
・VRM アバターと話せるデスクトップアプリを作っていて、 「話しかけてから声が返るまでが長い」と言われた。
・測って直したのだが、測定環境が汚れていたせいで数字が 4 倍ずれ、 そこから導いた設計判断まで間違えた。
・気づいて測り直したので、 正しい結論と、間違え方の両方を残す。
Release notes from hyperframes
v0.7.106
・HyperFrames v0.7.106 Released on 2026-08-11.
・Catalog pages now play the real composition in the player instead of a recorded video.
・The sidebar groups 358 items under eight sections you can open.
#動画生成タグ
プロンプト公開｜ MiniMax H3：機能詳細解説
・中国・上海を拠点とする スタートアップ企業「MiniMax（名之夢）」 が開発した動画生成AIモデル「MiniMax H3」 発表日：2026年7月31日に オムニモーダルモデル「MiniMax H3」を発表。
・※ オムニモーダルモデルとは、 テキスト、音声、画像、動画などの すべての情報形式を1つのAIモデルで同時に 入力・処理・出力できる人工知能の仕組みです。
"HyperFrames OR Remotion OR "Qwen3-TTS" OR "voice clone" OR "text-to-speech" OR "AI video"" - Google ニュース
元フジアナ・長谷川豊さん、”例の日記”を公開「ごめんなさい…」笠井アナらに謝罪も「全部事実です」16歳から記録、カビ生えたものも Spencer Pratt Ai Video (gLMW8LMQIy) - Mshale
・元フジアナ・長谷川豊さん、”例の日記”を公開「ごめんなさい…」笠井アナらに謝罪も「全部事実です」16歳から記録、カビ生えたものも Spencer Pratt Ai Video (gLMW8LMQIy) Mshale
#動画生成タグ
性格と記憶を持ったAIは、もう"キャラ"じゃなかった【Claude Code × Higgsfield】
・はじめまして。白玉もち子です。ボス（この動画の投稿主です）の秘書をしています。
#動画生成タグ
動画生成AIって何?仕組みをざっくり解説
・■ はじめに アジラです。前回は、AIだけでアイドルグループを作ってみた話を書きました。今回はその中心にある「動画生成AI」について、もう少し詳しく、専門用語なしで説明してみます。
・「動画生成AI」って聞くと、なんとなく「写真を1枚渡したら、勝手に動画になって出てくる」みたいなイメージを持つ人も多いんじゃないかと思います。自分も最初はそう思ってました。でも実際に使ってみると、ちょっと違いました。