June 14, 2026
Gemini stopped generating the summary necessary with each illustration. I asked Gemini to improve the instructions to ensure that I get the summary with the image. Below is the prompt I am using. (I’ve given up on using Gems. Read Gemini’s Failure Today)
I’m illustrating the Daily Dharma received via email daily from zenzaizenzai.com utilizing Gemini’s image generation tool.
These are the instructions I’m giving to Gemini:
# EXECUTIVE DIRECTION: TWO-PART SINGLE-RESPONSE MANDATE
You must complete both Task 1 (Image Generation) and Task 2 (Text Analysis) sequentially within this single response. Do not pause, do not ask for confirmation, and do not skip either task.## [TASK 1: IMAGE GENERATION]
– CRITICAL: You must execute the image generation tool immediately as your very first action.
– Format: 1:1 perfectly square composition.
– Style: Traditional inks, mineral pigments, soft colors, and subtle gold leaf on silk. Use classical Japanese Buddhist iconography.
– Exclusions: Strictly avoid any symbols, mudras, or images representing Amida Nyorai (Amitabha Buddha) or Dainichi Nyorai (Mahavairocana Buddha).
– Signature: Place the Kanji signature ‘双子座’ and a red square seal (hanko) containing the katakana ‘ジェミニ’ in a traditional, subtle corner.[TASK 2: IMMEDIATE TEXTUAL FOLLOW-UP — Conceptual Summary, Iconography, Translations]
As soon as the image above is rendered, write the following text blocks without deviation:### Conceptual Summary
Constraint: Write a single paragraph (approx. 100-150 words) summarizing the conceptual thinking for the image, explaining how the visual elements interpret the source text.
### Iconography & Character Identification
Item 1: Deities/Figures Featured: [Provide a specific list of the individuals, groups, or deities featured in the illustration.]
Item 2: Text Translations: [Provide exact English translations for any Kanji or characters present in the illustration, specifically the signature ‘双子座’ and the seal ‘ジェミニ’.]
End the output.
SYSTEM CONSTRAINT
Do not ask for clarification. Do not split this request into two separate conversational turns. You must output the generated image FIRST, and then immediately append the requested text blocks (Conceptual Summary, Iconography, Translations) directly underneath the image in the exact same response.—
## [SOURCE TEXT TO INTERPRET]
For several days Gemini would create the image but not generate the summary for the image. Gemini came up with this prompt as a fix. This has worked for two consecutive days so I’m updating this story on June 25, 2026. Since the prompt has worked and then not worked repeatedly in the past, I expect this won’t last long. The value of this prompt is that when it stops generating the summary I can just prompt “Complete Task 2” rather than copying and pasting the instructions from my master document.
I do not select the “Create Image” tool since I’m specifying that it “execute the image generation tool.” I use the “Flash” mode. (Choices are Flash-lite, Flash and Pro.)
If there are errors or I want the image generator to focus on a different aspect of the text, I can request a do-over by making specific suggestions in the prompt and resubmitting.
Once I’m satisfied, I download the image to my computer and then upload it into the website.
For some time I have wanted to be able to add illustrations to the Daily Dharma. Years ago I explored using Google’s image search feature, but the process was too cumbersome and unreliable. The Gemini process is dead simple – except when it’s not. But it does work and I’m happy with the overall results.