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.
Title cards for YouTube videos, documentaries, and short films — using Ideogram for text-led title cards is faster than animating one in After Effects (add animation in post with other tools).
Course thumbnails for Hahow / Udemy / Skillshare — text-led design keeps the title readable even at thumbnail size. Ideogram handles font contrast better than Midjourney.
A flowchart for educational content — GPT Image 2 is the most reliable tool for diagrams combining Chinese text, arrows, and numbered steps. Works well for social posts, slide decks, and book illustrations.
Turns your 5–8 most common customer questions into a single graphic — question, short answer, and an icon for each. Retains readers roughly three times better than a long text-based FAQ page.
A startup organizational chart with more design polish than a typical ugly PowerPoint org chart — for fundraising decks, careers pages, and company profiles.
Midjourney isn't the strongest logo tool, but it's more than enough for moodboard-level drafts — style references for a designer, or placeholders for a VC pitch deck.
A single moodboard image made before a design briefing — condensing the color, texture, lighting, and mood of a visual direction into one image, so a client can quickly confirm 'is this the feeling?'
A square-format e-commerce hero image — product centered, mood-setting background, negative space for adding text. Suited for Shopify collection pages, IG shop, and ad platforms.
Brand wrapping paper, website footer backgrounds, social post backdrops. Recraft's pattern feature guarantees the tile lines up perfectly, with no broken seams at the edges.
Ideogram 3.0's text-in-image is the most accurate of any current image model. Not just English, it renders Chinese, Japanese, and Korean precisely too, ready to use for logos, signage, and book covers.
What brand social accounts always lack is content that's funny without being cringe. GPT Image 2 produces four-panel comics with far better consistency than Midjourney: the same character looks identical across all four panels, the scene flows continuously, and you can even drop in Chinese dialogue.
PRO12The biggest headache for self-published authors on Amazon KDP, Readmoo, and similar platforms is the cover. One prompt produces three usable versions.
免費13Generates 12 mutually distinct logo sketches in one go — still using Fiverr for this is already behind the curve. Also gives you more leverage when pitching to a client.
免費14For travel bloggers, YouTubers, and newlyweds. Run a moodboard set to decide what to shoot, what to wear, and what kind of accommodation to book for the trip.
免費15An illustration mood similar to Makoto Shinkai or Kyoto Animation, suited for cover art, LINE stickers, and social media assets.
PRO16A standard ecommerce hero-image template — product plus subtle glow plus a minimal background, ready to publish as-is.
免費17A magazine-cover-like portrait style, suited to personal-brand avatars and portfolio hero images.