Google Gemini 新增音樂生成功能
💡Gemini generates music from text/images/videos – multimodal audio for creators unlocked
⚡ 30-Second TL;DR
有什麼變化
Gemini 應用程式現支援音樂生成
為什麼重要
此更新強化 Gemini 在創作 AI 的地位，吸引音樂家及創作者加入 Google 生態系統。它加劇了與 Suno 或 Udio 等生成音頻工具的競爭。
下一步行動
Update Gemini app and test music generation from an image prompt like 'guitar solo'.
關鍵要點
- •Gemini 應用程式現支援音樂生成
- •接受文字、圖像及影片作為輸入參考
- •可從多模態提示創作音樂
🧠 深度解析
背景與延伸：來自公開資料，非原文內容。引用 6 個來源。
🔑 增強重點摘要
- •Gemini app integrates DeepMind’s Lyria 3 model for music generation, producing 30-second tracks with lyrics and cover art from text prompts[1][2][3].
- •Supports multimodal inputs including text descriptions, uploaded photos, or videos to match mood and generate fitting music[1][3][5].
- •Lyria 3 enhances realism, musical complexity, user control over style, vocals, tempo, and automatically generates lyrics[1][2][3].
- •Features SynthID watermarking on all outputs for AI identification, plus detection tools for uploaded audio in Gemini[1][3].
- •Available globally to 18+ users in English, German, Spanish, French, Hindi, Japanese, Korean, Portuguese; rolling out from February 18, 2026[1][5].
📊 競品分析▸ Show
| Feature | Google Gemini (Lyria 3) | Suno | Udio | MusicGen (Meta) |
|---|---|---|---|---|
| Input Types | Text, image, video | Text | Text | Text, audio |
| Output Length | 30 seconds | Up to 4 min | Up to 4 min | Variable |
| Lyrics Generation | Yes, automatic | Yes | Yes | No |
| Watermarking | SynthID | Yes | Yes | No |
| Pricing | Free (Gemini Advanced?) | Freemium | Freemium | Open-source |
| Languages | 8 supported | Multi | Multi | English-focused |
🛠️ 技術深入
- Powered by Lyria 3, Google DeepMind’s latest generative music model, improving on prior versions for more realistic, complex tracks with natural flow and high-fidelity audio[1][2][3][6].
- Generates tracks with lyrics, instrumentals, vocals in multiple languages; users control genre, mood, tempo, dynamics, drumming style[1][2][3][6].
- Integrates Nano Banana for cover art; outputs exportable crisp audio with embedded SynthID watermark[1][2][3].
- Beta feature; no specific architecture details like parameters or training data disclosed in sources[1][3].
🔮 前景展望AI analysis grounded in cited sources
Expands Gemini's multimodal capabilities into audio, enabling custom soundtracks for personal use, YouTube Shorts via Dream Track, potentially integrating into apps like Google Messages; raises AI music detection needs with SynthID advancements[2][3].
⏳ 時間線
📎 來源 (6)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
- TechCrunch — Google Adds Music Generation Capabilities to the Gemini App
- engadget.com — Gemini Can Now Generate a 30 Second Approximation of What Real Music Sounds Like 204445903
- Google Blog — Lyria 3
- thurrott.com — Google Gemini Can Now Generate 30 Second Music Tracks
- workspaceupdates.googleblog.com — Create Custom Soundtracks with Lyria 3
- youtube.com — Watch
AI 週報
閱讀本週精選 AI 大事摘要 →
👉相關動態
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： TechCrunch AI ↗
每週 AI 簡報
每週一封，可隨時退訂。