💡 AI 模型在英文提示詞下表現最佳。因此,提示詞本文以英文呈現。使用英文輸入可獲得更準確、更詳細的回應。 找到下一部最愛的電影,不該花 45 分鐘滑來滑去。這些提示詞讓 AI 變成你的私人影評——而且真正了解你的品味。經過四大主流 AI 模型實測,給你真正的推薦,而非千篇一律的十大排行榜。
提示詞
建立個人化的電影喜好檔案
Here are my 10 favorite movies of all time: [List 10 movies you love] And here are 3 movies I hated: [List 3 movies you disliked] Based on this, give me: 1. My "viewer DNA" — the themes, pacing, and storytelling styles I gravitate toward 2. 10 movies I've probably never seen that match my profile 3. 3 directors I should explore and why 4. 1 genre I'm sleeping on that I'd probably love 5. A "stretch pick" — something outside my comfort zone that might convert me
進階技巧
用幾個字說明你為什麼喜歡或討厭每部電影。「喜歡《寄生上流》——階級評論的部分」比只寫片名能讓 AI 更精準判斷你的喜好,效果好五倍。
已測試 Mar 15, 2026
根據當下心情找到合適的電影
I'm in the mood for something [describe your mood: cozy, thrilling, mind-bending, heartwarming, dark, etc.]. I'm watching with [alone / partner / family / friends]. Time available: [90 min / 2 hours / open-ended binge] Platforms I have: [Netflix, Hulu, Disney+, etc.] Give me: 1. 5 perfect picks ranked by match strength 2. For each: title, year, one-line hook, and where to stream it 3. One wildcard pick I wouldn't think to search for 4. A backup option if my group vetoes the top pick 5. The ideal snack pairing for the vibe (just for fun)
進階技巧
提到你最近看了什麼、喜不喜歡。AI 會利用近期觀影經驗,避免推薦太相似或太不同的作品。
已測試 Mar 15, 2026
找到值得一口氣追完的電視劇
I need a new TV series to binge. Here's what I've loved: [List 5 TV series you enjoyed] Preferences: - Episode length: [20 min / 45 min / any] - Completed or still running: [preference] - Seasons: [1-2 / 3-5 / don't care] - Content I avoid: [violence level, topics, etc.] Give me: 1. 5 series recommendations with why each matches my taste 2. The ideal starting episode if the pilot is weak 3. Expected binge time (total hours) for each 4. Which ones have satisfying endings vs. cliffhangers 5. A "one-season wonder" I can finish in a weekend
進階技巧
說明你想看已完結還是連載中的影集。沒有什麼比追到一半發現被砍、留下懸念更痛苦的了。
已測試 Mar 15, 2026
規劃主題電影之夜
I'm hosting a movie night with [number] people. Theme idea: [e.g., 80s nostalgia, horror comedy, Oscar snubs, foreign thrillers] Group age range: [e.g., 20s-30s] Vibe: [casual fun / cinephile deep-dive / family-friendly] Time for: [1 movie / double feature / triple feature] Plan my movie night: 1. Movie selection(s) with runtime and content warnings 2. Viewing order if multiple films (and why that order) 3. Discussion questions for each film 4. A themed drink or snack idea 5. A 30-second intro I can read before pressing play 6. A backup movie if the group wants something lighter/darker
進階技巧
告訴 AI 你們之中最挑剔的觀眾是誰。如果有人討厭恐怖片,整個計畫都得調整。AI 無法針對它不了解的群體偏好做最佳化。
已測試 Mar 15, 2026
深入了解一個全新類型
I want to explore [genre] movies but I'm completely new to the genre. What I usually watch: [your normal genres] What intimidates me about [genre]: [e.g., too slow, too violent, confusing] Time commitment: [start slow with 1 film / give me a full curriculum] Create my genre introduction: 1. The perfect "gateway" film and why it works for newcomers 2. A 5-film curriculum from accessible to challenging 3. Common misconceptions about this genre 4. Key directors and their signature styles 5. The vocabulary I need (terms, movements, sub-genres) 6. One film to avoid as a first exposure and why
進階技巧
坦白說出讓你卻步的原因。說「我覺得黑白片很無聊」並不膚淺——這能幫助 AI 先推薦彩色片,之後再慢慢帶你入門經典作品。
已測試 Mar 15, 2026
基於實際測試結果 — 非假設推測。 查看測試方法
Claude Sonnet 4
最深入的品味分析,最擅長發掘真正冷門的電影。能辨識你偏好中的微妙模式,推薦其他模型忽略的作品。
最佳品味分析Gemini 2.5 Pro
串流平台可用性、片長和結構化觀影課程最準確。提供精確的集數和誠實的跨季品質評估。
最佳事實細節GPT-4.1
最擅長團體推薦、派對規劃和根據心情選片。自然地處理討論問題和主題搭配等創意元素。
最佳社交觀影Grok 3
擅長推薦小眾、邪典和被低估的電影,具有超越演算法推薦的強烈個人觀點。
最佳邪典選片告訴 AI 你討厭什麼,比告訴它你喜歡什麼更有用——知道你受不了慢節奏的片子,就能立刻排除一半的推薦選項
預先說明你有哪些串流平台,這樣每個推薦都是你今晚就能看的
要求 AI 說明每個推薦的「原因」——如果理由讓你無感,這部片大概也不會合你胃口