Nano Banana Pro: compare prompt following across several image tasks

Choose the image task and preserve the reference

The November 2025 tutorial demonstrates realistic and cartoon images, thumbnail face replacements, Google Maps-based scenes, actor lineups, infographics and editing inside Higgsfield. These are different jobs with different checks. Decide whether you want a new composition or a controlled change to an existing reference before generating.

The original product destination is https://higgsfield.ai/nano-banana-2-intro. Despite that URL’s wording, the supplied video names Nano Banana Pro. Preserve the original resource rather than silently substituting another model or destination. The description’s unlimited 4K for 365 days is a historical offer, not verified current access.

1. Follow the image examples at 0:31 to identify the intended style.
2. Keep the original thumbnail or source photo before editing.
3. Describe what should change and what should remain visually consistent.

Compare editing and location examples with the input

Thumbnail face replacement begins at 1:29 and Google Maps photos at 2:48. Inspect the generated result against the actual input, including facial identity, perspective and the scene details that matter to the task. A realistic appearance alone does not establish that a place or person has been reproduced accurately.

Actor lineups and video animation follow at 3:53. Select the still before judging its animated version, then watch for changes introduced by movement. Keep the generation stages separate so a defect can be traced to the image or the video.

1. Compare the edit with the requested change and the preserved reference.
2. Inspect the location-based scene for altered details and composition.
3. Watch the whole animated output for consistency with the chosen still.

Read the infographic and inspect the final size

The infographic example starts at 6:00, followed by 4K images and celebrity photos at 7:56. Read every line of an infographic rather than relying on a general impression of crisp text. Check wording and layout at the size the audience will see.

The title asks whether this is the best prompt-following model, but the description supplies no controlled ranking across all alternatives. Use the examples to judge specific instructions and outputs. Keep any historical offer language separate from the quality assessment, and preserve the selected version so later edits can be compared with a clear starting point.

The takeaway

Assess each task against its input and intended use: accurate edit, coherent animation or readable infographic, rather than a single overall quality claim.