1328 hand-tested AI prompts across image, video, and music — each with full context, variables, and a sample output. Content is authored in Traditional Chinese; the UI here is English.
Sora 2 responds best to this kind of writing for 'physical realism' — using a dedicated `Physics:` section to call out surface tension, …
The key to image-to-video is 'only describe the change, don't redescribe what's already in the frame' — this prompt uses five sections, `Transformation/Li…
Write the audio as a bulleted `Audio requirements:` spec — the bag-impact sound 'synced to the motion,' breathing, …
The hardest part of a two-person dialogue is 'keeping both characters stable' — this prompt uses `[Character Cameo: …]` to separately lock in the persona of the host and the guest, …
Mood pieces most easily turn into a pile of adjectives — this one instead uses five fields, `Mood / Atmosphere / Detail…
One prompt runs through four shots: a low-angle wide shot establishing the giant figure, a dolly-back retreat, a cut to handheld footage following the fleeing crowd, …
The tension in sci-fi action comes from 'the camera suffering along with the characters' — handheld FPV riding on the shoulder, a whip-pan catching a terrified face, …
Breaks a piece of content into N key points, laid out as a Traditional Chinese infographic card with a title, bullet points, and labels, all rendered correctly.
A signboard for a Taiwanese night market food stall, with a vertical-text menu, prices, and bold hand-lettered type — unmistakably Taiwanese with no garbled characters.
A repeatable Traditional Chinese cover-card system with a fixed color scheme and type hierarchy — swap in a new headline and the whole account's visual identity stays consistent.
A fill-in-the-blank Traditional Chinese poster template: define the layout hierarchy before adding text, and keep a safe zone around the QR code and logo so the background doesn't overwrite them.
Composites two portraits into one vintage Polaroid — warm, faded tones, a white border, and a handwritten Traditional Chinese note — with special handling for Asian facial consistency.
Uses a real product photo as reference and locks the brand color via hex code, describing a white-background hero shot plus lifestyle shots and multiple angles in a single natural-language prompt.
Writes realistic portraits using camera-and-lighting language, locks in the likeness with a reference image, and mass-produces a stylistically consistent set of personal brand photos across different poses.
Uses a precise dictionary of Taiwanese elements to realistically generate night-market and temple-front scenes, with explicit exclusion lists to avoid a generic pan-Asian mishmash.
Midjourney v7 uses 85mm backlit portrait phrasing plus --sref to lock a brand's color tone, letting instructors, real estate agents, and solo creators shoot a whole visually consistent set of headshots in one go.
Midjourney v7 uses a precise dictionary of Taiwan-specific elements (vertical neon signage, red lanterns, corrugated-metal arcades, scooter swarms) to avoid generating scenes that read as Japan or Hong Kong instead — for local-revitalization and food/tourism material.
Ideogram 3.0 creates front labels and packaging for food and handmade goods, rendering the brand name in Chinese and English side by side plus precise fine print for volume and ingredients — built for Taiwanese small-business food and craft brands.
The fix for the most common pitfall among Chinese-language users. Put the style description in English in the Style field, put Traditional Chinese lyrics in the Lyrics field, and keep each section to 4-6 lines so the AI doesn't rush or drop lines. Includes three swappable scenarios: wedding songs, class songs, and shop theme songs.
A faceless-channel music formula for sleep, studying, and late-night companion content. Match BPM directly to physiological state — 50 for sleep, 85 for studying, 100 for driving — then use an exclusion list (explicitly stating what you don't want) for far more accurate results than vague mood descriptions. Includes a temple-chant healing style.
Uses Udio's strength in layered vocals and authentic-sounding acoustic instrumentation to generate a Mandarin acoustic duet with male and female leads and three-part vocal harmony in the chorus.
Uses Udio's realistic instrument textures to build a three-piece kit — an intro hook, a transition sting, and a closing outro — all built on the same motif to keep the show's identity consistent.
Feed the AI the background of a matter and it produces a draft following Taiwan's official document format — Subject, Explanation, Action — automatically applying standard bureaucratic phrasing and adjusting tone for upward, lateral, or downward correspondence. The final section forcibly flags regulations and figures that "need human verification"; this is positioned as a draft, not a final document.
Bilingual Chinese-English localization with one more layer than Google Translate. Specifies Traditional Chinese as used in Taiwan plus Taiwan-specific idioms, keeps names, brand names, and technical terms from being mistranslated, and delivers both a literal version and a localized, polished version in one pass so you can compare and pick. You can also feed it style samples so it learns your company's tone. Suited to marketing copy and product descriptions for foreign firms, freelancers, and cross-border e-commerce in Taiwan.