Advanced Two-Step Prompting Strategy for Nano Banana Pro Thumbnails

Actually, there's a slyer way to use this.

First, have it systematically summarize
→ Then based on that, have it create a deliverable / dig deeper into the topic
Just doing this makes the quality of the output really different.

Why? Because initially, have AI create a blueprint in its mind
with 'terminology', 'judgment criteria', 'prohibitions', 'templates', 'priorities', etc.,

then the next instructions will output based on that blueprint, so it's less likely to deviate.

For example, if you want to have Nano Banana Pro create a thumbnail,

① First, have GPT-5.2
'research thoroughly and systematize the best practices for creating thumbnails with Nano Banana Pro. Include common failures as well.'

② Then,
'Based on the current system, create a prompt for making a YouTube thumbnail with Nano Banana Pro for the theme {argument name="テーマ" default="◯◯"}. The video content is something like this, and I want to include {argument name="文言" default="こういう文言"}'
like this, incorporate your requirements and have it create the prompt.

Then put the completed prompt into Nano Banana Pro. It becomes much easier to get the desired output.

If you just ask 'make it' every time, AI also produces on the spot, so there are hits and misses, but
if you have it create a map (system) first, it's powerful.

Moreover, this method
works for side business strategies, sales documents, video scripts—everything. The versatility is amazing.

Before having AI do work, it's really recommended to have it organize how to make good deliverables, and then have it think based on that.

By the way, the thumbnail made this time is the third one.
The first time was mediocre, so I had it systematize how to create attractive YouTube thumbnails and then create the prompt.

Machine-translated — the original is what produced this result.

✨ Generate Save Copy prompt Share Original post Open in Hitboard