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.
Short, spec-style phrasing — focal length, light position (backlight creating a halo), camera position (static), motion (slowly...
The phrase `flickering streetlamp` supplies the light source, flicker motion, and a period feel all in one; `rai...
Studio lighting terms are used as literal parameters — key light and fill light each get a specified direction and gear (window left / bounc...
The hard part of a loop is making the start and end match — using the three columns `First frame / Last frame / Mi...
Native audio is the thing models most often bury under auto-generated music — start by setting the spec with `Native audio. 48kHz.`...
Feed in a textbook chapter, lecture notes, or meeting materials, and get three items generated at once: flashcards to help you memorize, a self-quiz to test yourself, and a speaking outline to help you explain it out loud. Good for certification exam prep, study-group facilitation, and new-hire training. It has a built-in anti-fabrication rule that says "organize only from the provided material"; for long material, feed it in batches and generate one pack per batch, then merge — that's the most reliable approach.
Turns budget, walking pace, traveling with elders, and rain contingencies all into hard constraints that trigger a full re-plan if violated, forcing the AI to build an itinerary you can actually walk through, with a rain backup and daily cost subtotal for every day, and a constraint-check pass run after planning is done. Good for family trips or traveling abroad with elders. Note: the AI often gets business hours and ticket prices wrong — every such item is forcibly flagged "confirm before departure," and you should always check the official website yourself before you go.
Doesn't let the AI just hand in its answer as-is: it first tags the source type of every key sentence, then hunts for counterexamples to attack its own claims, and finally grades confidence level item by item with suggested directions for verification. Good homework before researching, writing a report, or making a decision. Honest note: this isn't the same as actual fact-checking (the model may not be able to search the web) — but it forces uncertainty out into the open and noticeably reduces confident-sounding nonsense.
Feed in your shop name, business hours, price range, and tone persona, and get four LINE Official Account template sets generated at once: a price-inquiry guide, a booking confirmation with a fully-booked alternative, a two-part complaint reply, and a closing auto-reply. The tone stays consistent throughout, and each message is kept to a length readable in one message. Double-check prices and hours against each template before pasting them into your backend — the AI doesn't know your shop's actual real-time situation.
Report a bug in this format and the AI's fix hit rate jumps noticeably: give all four parts at once — the raw symptom, a minimal repro, expected behavior, and environment — and force the AI to diagnose first, ask for more information when it's insufficient, and never guess-and-change blindly. Suited to any scenario where you're asking ChatGPT or Claude to fix code. Note: the AI's diagnosis is reasoning, not an actual test run — after the fix, be sure to actually run through the included verification checklist.
Feed in a lease, employment, or outsourcing contract, and it scans clause by clause from your side of the deal: a plain-language translation, worst-case scenario, risk grading, and negotiation angle for each, plus three questions to bring to a professional at the end. Note: this is a prep tool for understanding a contract, not legal advice — for high-value contracts or ones involving major rights, consult a lawyer, and it's safest to treat the scan results as homework to do before meeting one.
Paste in your health checkup numbers and get each one translated into plain language, with flagged (out-of-range) items collected together, plus a concrete list of questions you can read straight off at your follow-up appointment. Solves the pain point of "seeing the flagged numbers, not understanding what they mean, and not knowing what to ask in a 3-minute consultation." Not medical advice — abnormal values are always for your doctor to interpret; the AI's role is limited to translation and coaching you on questions, and it is barred from making a diagnosis.
Generate a vertical-format price board for a bubble tea shop or food stall: each item and price locked with full-width quotation marks, character counts noted per line, and safeguards against simplified-character variants. Built for boards headed straight to print. Honest caveat: the more items, the higher the error rate — beyond ten items, split into two generations; for a one- or two-character mistake, just regenerate, or ask the model to fix only the wrong lines.
Generate a step-by-step SOP diagram for staff training in one pass: the title and every step's text locked with full-width quotation marks, character counts noted per step, and step numbers in Arabic numerals to reduce the error surface. Good for follow-the-steps scenarios like opening procedures or cleaning protocols. Limits: keep steps to six or fewer, with roughly ten characters per line for the most stable results; the illustrated action occasionally doesn't match the text — just ask it to redraw that one panel.
Two key priorities for generating event tickets and staff passes: locking the event name and date in quotation marks (with character-count declarations and safeguards for characters like '證/号'), and clearly reserving a pure white blank zone for a serial number or QR code to be added later. It specifically adds a 'no fake barcodes or serial numbers' rule — the model loves to invent a fake barcode on its own initiative, and you don't want to find out it can't be scanned only after printing.
The most painful part of e-commerce lifestyle shots is 'each of the four images looking like it came from a different shoot': generating them in four separate passes almost always shifts the product's appearance and lighting. This approach instead generates a single 2x2 four-panel grid in one pass — since all four panels share the same generation, consistency is far more stable than generating separately. The trade-off must be stated plainly: each panel is only a quarter of the full image's resolution, small text on product labels may come out blurry, and you'll need to upscale or regenerate individual panels when you need a larger image.
Handles the two big requirements for wrapping paper and bag patterns at once: tileability, and staying unobtrusive as a background so a logo doesn't get lost in it. Achieved via low density plus even negative space plus low contrast to leave breathing room for a logo, and by banning directional lighting/shadow and perspective to preserve the tiling. Honest limitation: generative models don't guarantee true seamlessness — check the edge seams in image-editing software using an offset test, patch locally wherever there's a visible seam, and don't expect it to be perfect in one shot.
The old problem with generating recipe step photos separately: the countertop, cookware, and lighting jump around from shot to shot, and readers spot the inconsistency immediately. This approach generates a single grid in one pass to lock consistency, then constrains ingredient state progression with the rule 'each panel must be a plausible continuation of the one before it.' Limitation: the more panels, the smaller each one gets — split into two images past six panels, and the second image needs to repeat the same countertop and lighting setup to connect properly.
Weaves the couple's names, where they met, and how long they've been together into a Taiwanese Hokkien wedding love song: the verse tells how they met, the chorus is a vow to hold hands together, and the bridge closes with blessings from parents and guests — usable for the wedding entrance or a relationship retrospective video. Hokkien pronunciation occasionally goes off — the context section includes tone error-proofing and fixes, and a few regenerations usually get it right.
15 seconds is only enough for people to remember one thing: the brand name. This uses a single hook melody repeated three times, with the brand name appearing at the beginning, middle, and end, quotation marks to lock the wording and prevent mis-singing, and a clean ending landing on the brand name. Suited to in-store announcements, short-video openers, and ad bumpers. Trim the duration with the web app's slider, or cut it yourself after generation.
Breaks daily routines like brushing teeth or putting away toys into three steps and fits them into a children's song with a repeating sentence pattern: once the child remembers the melody, they remember the routine. The chorus uses positive encouragement like "I'm the best," and ending with the child's nickname boosts engagement. Good for bedtime routines or cleanup time — play it and let the child sing along. Switching topics just means changing the three step variables.
Generates long-playable, vocal-free ambient background music for cafés: the prompt explicitly avoids dramatic peaks and keeps energy level consistent start to end, so tracks can be chained into longer sessions using the Extend feature. Honest caveat: Udio's download/export is currently disabled, so output can only be played within the platform — this is best used now to dial in a style you'll export once downloads return.
A podcast's intro and outro need to sound like they're from the same show: this gives you a shared style description plus two version blocks, and makes it a hard rule that "the outro is cut from an Extend of the intro" to ensure the melodic motif is genuinely shared. The 15-second intro ends on an open, unresolved feeling leading into the opening line; the 30-second outro closes on a sense of resolution — standard broadcast grammar.
Hailuo 02 is most stable with a short, specific single sentence — subject (astronaut) + setting (neon-lit rainy Tokyo alley) + action …