In partnership with
LUXE PROMPTING
GEMINI IMAGE WORKSHOP
Gemini: Turn a Rough Sketch Into a Clear Explainer
A polished diagram can still explain nothing. Before asking Gemini for an infographic, draw three boxes and two arrows. That rough layout can make your intention clearer than a paragraph of style adjectives. Here is a complete workflow for turning a simple creative process into a useful, readable visual.
Advertisement
Is Your Training Data Actually Model-Ready?
If you're fine-tuning a speech model, you've probably hit this wall: DNSMOS gives you a score, but it doesn't tell you whether the data behind that score is actually right for your model.
Treat it as a pass/fail gate and you'll end up training on audio that looks clean on paper but drags down real-world performance—while good source data gets tossed for no reason.
Voices' CTO DJ Jalali (with the team's senior audio and voice data engineers) just published a free white paper that breaks down the four-step calibration framework they use internally to set model-specific quality thresholds instead of trusting the raw DNSMOS number. It also covers where DNSMOS breaks down and how Voices validates audio for custom datasets at scale.
The Short Version
- Draw the reading order before generating the design.
- Supply approved labels instead of asking the image model to invent facts.
- Check the arrows and sequence as carefully as the spelling.
Start with a deliberately plain sketch
Draw three boxes stacked vertically. Label them BRIEF, CREATE, and CHECK. Add one downward arrow between the first and second box and another between the second and third. Leave a space for the title at the top. A phone photo of this sketch is enough to communicate the intended arrangement.
Google documents both image editing and rough-sketch refinement for Gemini image models. We are using that capability as a layout instruction. We are not asking the model to research a technical process or decide which steps belong in it.
Copy this sketch-to-explainer prompt
Use my uploaded sketch as the layout reference for a clean portrait explainer graphic. Keep exactly three vertically stacked steps and two downward arrows. Use a white background, dark charcoal text, and restrained purple accents. Give each step generous space and a small, simple line icon. Do not add decorative objects.
The title is "A CLEARER CREATIVE WORKFLOW". Step 1 is "BRIEF" with the sentence "Choose one result and one audience." Step 2 is "CREATE" with "Make one version from the approved brief." Step 3 is "CHECK" with "Compare the result with the brief." Render each supplied phrase exactly once. Use readable sans-serif lettering with the step labels clearly larger than their sentences.
Keep the arrow direction top to bottom. Do not create extra steps, side branches, percentages, claims, logos, footnotes, or other text. Keep all lettering away from the edges. The goal is a diagram that reads clearly on a phone.
Review the logic before the decoration
Count the boxes. Count the arrows. Follow the route with your finger. A fourth panel or a backward arrow changes the meaning, even when the colors look excellent. Compare the entire graphic with the sketch before judging the icons.
Next, read every line out loud from the image. Watch for repeated words, missing punctuation, and text moved under the wrong step. If the diagram needs small print to make sense, shorten the copy or split the explanation into two graphics.
Repair the weak part
If the second step is crowded, request only a spacing adjustment in that panel. If an arrow points sideways, identify that arrow and its intended destination. Keep the approved words and other panels fixed in the revision brief, then inspect them again because unintended changes remain possible.
For an explainer that will change often, use the AI result as a visual concept and rebuild the labels in a slide or design editor. Editable wording is more useful than repeatedly regenerating a graphic whenever a step changes.
Make it useful outside the image
Try this with your own onboarding flow, content checklist, or workshop exercise. Supply the facts yourself and treat the model as a visual collaborator. The finished graphic should clarify the process, not become the only place the instructions exist.
- Sequence: three approved steps in the right order.
- Labels: every sentence matches your source text.
- Legibility: readable on a phone without enlargement.
- Access: repeat the essential instructions in the accompanying text.
Source Desk
Official sources checked September 22, 2026. Recheck current details before acting.
Google: Gemini image generation, editing, and sketch refinement
A clearer brief. A more useful result.