Writing prompts from scratch is the slowest part of working with AI image and video models. The fastest shortcut I have found is to start from a prompt library that shows the actual output next to each prompt, copy the one closest to what you want, and edit from there.
Here is what to look for in a free prompt library, plus the two I use most for Gemini Nano Banana (images) and Seedance (video).
What makes a prompt library actually useful
- Real previews. A prompt without its result is just a guess. You want to see the image or clip it produced.
- Copy-ready wording. Full prompts, not keyword soup, so you can see how style, lighting and constraints are phrased.
- Filters that match how you work. Aspect ratio (16:9 vs 9:16) for video, and categories like portraits, product shots or restorations for images.
- Free to browse. You should be able to study and copy prompts without paying first.
Nano Banana (Gemini image) prompts
The Soutine Nano Banana prompt library has 1,400+ free image prompts covering photoreal portraits, product staging, anime, interiors, photo restoration and surreal concepts. Each card shows the result, and you can copy the prompt or open it straight in a generator.
Patterns that show up again and again in the prompts that work well:
- Say what must stay the same when editing a photo ("keep the face, pose and background unchanged").
- Name the light ("soft window light from the left", "golden hour rim light").
- Give the camera a job ("85mm portrait, shallow depth of field").
Seedance (AI video) prompts
For video, the Seedance prompt library pairs about 66 prompts with the clips they generated. You can hover to play, filter by 16:9 or 9:16, and copy the exact shot description.
Three habits from those prompts that noticeably improve clips:
- Name the camera move ("slow dolly-in", "orbit around the subject").
- Put one beat in the middle of the clip, e.g. a turn, a reveal or a splash, so the motion has a purpose.
- End with hard constraints like "no text" or "seamless loop".
A simple workflow
- Browse a library and pick the closest result to your idea.
- Copy the prompt and change only the subject first.
- Adjust style and lighting words one at a time so you learn what each change does.
- Save the versions that work. Your own mini library compounds quickly.
Wrapping up
You don't need a paid course to get better AI images and videos. Studying prompts next to their real outputs teaches you faster than anything else. Both libraries above are free to browse and copy at Soutine AI. Generating new images or clips there uses credits, but the prompts themselves work as a reference for any model.
What prompt patterns have worked best for you? Share them in the comments.
Top comments (0)