Diverse images¶

• --num 4 makes four images at once: four requests run in parallel, so four take about as long as one.
• On a simple subject, plain --num 4 returns four near-copies.
• --deltas or --diverse makes each take look different.

genimg "a SINGLE minimal fox logo, NOT a grid" \
  --model gdm:nb2 \
  --num 4 \
  --grid

Name the styles: --deltas¶

• You choose the look of each take, on any provider and any subject.

genimg "a minimal fox logo, NOT a grid" \
  --model oai:gi2.5-flare \
  --num 4 \
  --deltas "line art, block print, brush stroke" \
  --grid \
  --open

Pass one delta fewer than --num, comma-separated or as --deltas @styles.txt with one per line.

Let genimg pick: --diverse¶

• No styles in mind? genimg picks a different look for each take, and each run differs.

genimg "a SINGLE minimal fox logo, NOT a grid" \
  --model gdm:nb2 \
  --num 4 \
  --diverse \
  --grid \
  --open

The built-in pool suits illustrations and logos.

Let the model pick: --mode batch (Gemini)¶

• One request asks Gemini for the whole set, so the model sees every take and makes them differ as a set, instead of genimg adding a style to each.
• The model decides how many images to return, so you may get fewer than --num; the default parallel mode always returns --num images.

genimg "a minimal fox logo, NOT a grid" \
  --model gdm:nb2 \
  --num 4 \
  --diverse \
  --mode batch \
  --grid \
  --open

Rules of thumb¶

• Both need --num 2 or more.
• Parallel mode runs up to five requests at a time (two for codex:image), and keeps the images that succeed if one fails.
• --mode batch is Gemini only. On OpenAI genimg rejects it, because batched takes come back as near-duplicates.
• --deltas and --mode batch do not combine.
• --grid --open opens the set in the grid, with each card's delta under it. A --mode batch set has no per-card deltas; its info panel says model-diversified instead.
• Found a winner? Edit it with --input (Input and reference images).