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.
PRO01A distressed vintage Wild West poster in a tall 23 composition, faithful to the reference image: an aged beige paper sheet pinned at the four corners against a pure black background, with torn irregular
PRO02An ultra-detailed surreal cosmic portal scene, faithful to the reference image: a vast dreamlike cloudscape suspended in a black void, with a sequence of large transparent rectangular planes receding
The 24 line icons your brand UI, app, or website is always missing. Recraft's strength is running a whole set in one stroke style, and you can export straight to SVG for the front end when it's done.
For enterprises, agencies, and e-commerce sites where licensing matters most. Firefly 4's training set is commercially safe — no IP landmines, ready to ship in a campaign.
The hero visual for the first page of a VC raise, client pitch, or annual report. Firefly's abstract business-feel imagery has clear commercial licensing that enterprise legal will clear.
The classic "product emerging from smoke" shot for fragrance, skincare, and spirits ads. Luma's fluid physics is the most realistic of any video model, so the smoke never looks like CGI.
Udio outperforms Suno on lyrical storytelling and vocal texture. Brand-story shorts, documentary songs, podcast theme music — one prompt handles it all.
The signature K-pop hook formula of "three beats + a high-note burst." TikTok and Reels have been eating this up for ages. Udio's Korean pronunciation is steadier than Suno's, and you can switch to English too.
For exhibition, talk, and event posters, Ideogram 3.0 is far more stable than Midjourney: headline, subhead, and date/location all render cleanly across three tiers, and it handles the hierarchy on its own.
An Instagram carousel needs different text on all 10 slides but a consistent overall look, exactly where Midjourney is weakest. Ideogram unifies the headlines, body copy, and icons into one layout, ready to post as a 10-slide carousel.
Pika 2.0's Morph effect lets character A turn seamlessly into character B. The transformation transition is the hottest content format on IG Reels and TikTok in 2025-26.
Pika 2.0's Explode effect slices, bursts, and scatters food in slow motion. For the first 3 seconds of a restaurant, delivery, or food-brand ad, nothing is more eye-catching.
Hailuo 02 beats other models by two tiers on animal fur dynamics, muscle movement, and eye detail. For pet accounts, brand mascots, and documentary B-roll.
A 15-second jingle small business owners can produce themselves for radio or elevator ads. Drop in your brand name, land the hook at the 10-second mark, and one pass gives you something ready to drop onto a podcast intro.
The real bottleneck in vlogging isn't editing, it's the music. This prompt generates a purely instrumental cinematic underscore whose emotional arc cuts beautifully against a vlog's pacing.
A narrative song for family oral histories, tribute videos for elders, and marketing aimed at older communities. Taigi plus a folk structure — Suno's Taigi pronunciation is steadier than you'd expect, and the hook is catchy.
The first 5 seconds of an e-commerce ad decide whether someone scrolls away. Seedance 2.0's multi-shot single take strings unboxing, use, and reaction into one cut-free video, with off-the-charts trust.
The "subject walking slowly in profile" shot fashion TVCs lean on constantly. Veo 3.1 handles fabric physics, hair dynamics, and gloss reflections more reliably than Sora — drop it in as the hero shot for the first 5 seconds of a 15-second ad.
The interviewee close-up for brand documentaries, internal-training onboarding, and testimonial walls. Veo 3.1's natural eye contact and lip-sync accuracy are good enough to carry the lead with no cut-and-paste.
GPT Image 2's most underrated ability is making infographics. Its Chinese character rendering is 10x more reliable than Midjourney's; feed it data plus structure and it produces an infographic ready for an annual report, a deck, or LinkedIn.
The biggest pain point with Chinese posters is that every image model garbles the characters. GPT Image 2 currently renders Traditional Chinese better than anything else, and this prompt is built for "Chinese headline + layout + zero errors," print-ready out of the box.
You have one generic resume and want to apply to 10 companies at once. Claude reorders and re-emphasizes it for each job description without changing any facts.
Not a casual 'take a quick look at this code' request. One command yields a full review of the PR across four dimensions: scope/impact, security, performance, and readability.
For B2B client proposals, corporate anniversary and annual report videos, and similar corporate footage — the crucial first 3 seconds. Lets viewers immediately register that this is a company that takes its business seriously.