GPT Image 2.5 prompts
What Sunburst and Flare each do, how prompting changed from GPT Image 2, and four full prompts re-run side by side.
AI Model Prompt Libraries
Choose the tool you use, then browse public examples labelled with that model. Shared model-specific syntax and parameters stay intact.
Browse by creative use case →GPT Image 2 prompts
Free GPT Image 2 prompts, each shown with the image it made — portraits, posters, infographics, UI screenshots and product shots. Copy any prompt in one click.
Nano Banana prompts
Free Nano Banana (Gemini image) prompts, each shown with its result. Copy prompts for portraits, 3D figurines, edits, posters and photo restyles — free.
Nano Banana 2 prompts
Free Nano Banana 2 prompts paired with public creator results and their shared requests. Copy and adapt examples for figurines, portraits, posters and photo edits.
Midjourney prompts
Free Midjourney prompts paired with public creator images: cinematic portraits, posters, illustration styles and sref codes. Copy the shared prompt and parameters in one click.
Seedream prompts
Free Seedream prompts paired with public creator results: portraits, fashion, posters and photorealistic scenes. See the source and copy the shared prompt for each image.
Seedance prompts
Seedance 2.0 prompts and video prompt examples from public creator clips, with camera, lighting and action notes you can copy, study and adapt.
Veo 3 prompts
Free Veo 3 prompts for Google's AI video model: public creator clips and frames paired with their source, model label and shared prompt.
GPT Image 1 prompts
Browse creator examples recorded as GPT Image 1, with images and attributed source text. Inspect reference requirements before adapting a shared prompt.
For sources carrying an older recorded label, browse the GPT Image 1 library without relabelling the examples as GPT Image 2.
Compare shared prompt workflows
Compare the instructions creators actually shared before moving a prompt between tools. The table counts published examples with each recorded model label, mentions of reference images, and parameter strings such as --ar or --sref. These are observations about this catalog’s text, not tests of model capability, quality, availability or reproducibility. A mention does not establish that the original input image or all settings were shared.
| Library / recorded label | Published examples | Reference-image mentions | Parameter mentions |
|---|---|---|---|
| GPT Image 2gpt-image-2 | 2094 | 286 | 83 |
| Nano Bananananobanana-pro | 465 | 85 | 11 |
| Nano Banana 2nanobanana-2 | 284 | 64 | 19 |
| Midjourneymidjourney | 118 | 3 | 55 |
| Seedreamseedream | 88 | 20 | 6 |
| Seedanceseedance | 58 | 6 | 0 |
| Veo 3veo | 1 | 1 | 0 |
| GPT Image 1gpt-image-1 | 39 | 10 | 2 |
Match the task first: a reference-photo edit and a text-only scene need different input context. Keep the subject, composition and lighting instructions when adapting an example, then translate unsupported syntax into the controls of the tool you use. Open the creator source when the prompt relies on an image or setting you cannot see. Smaller collections provide fewer examples to inspect; they do not prove that a model is better or worse. This comparison covers the existing library labels and introduces no new model landing pages.