GPT Image 2.5 prompting guide: 8 rules from OpenAI's official examples, Flare runs 2x faster than Sunburst
Fun_Walk_4965 · reddit · 2026-09-10
A Reddit user published a tested prompt library for GPT Image 2.5 (open-sourced on GitHub), distilling OpenAI's official prompting guide into 8 rules:
- Name the deliverable first — the model picks composition defaults from that word
- Format doesn't matter, maintainability does
- Describe only what's visible: material, light source, color, medium
- For people, specify framing, gaze and hands
- Put on-image text in quotes with position and typeface; spell unusual words letter by letter
- For edits: one line for the change, one line for what stays
- Number reference images and give each a job (subject / outfit / palette)
- Iterate by feeding output back in, one change per turn, repeating constraints
Timing notes from the author's own runs: Flare and Sunburst take identical prompts, Flare is about twice as fast as GPT Image 2, while Sunburst is worth it for fine detail and dense text. The post includes full prompts and side-by-side outputs for a photoreal sailor portrait, a coffee-machine process infographic, and a streetwear ad with exact tagline text.
Related event: GPT Image 2.5 Prompting Guide: 8 Rules and 22 Workflows(2 posts)→
More from Multimodal
- Fish Audio ships open-weight Drama 3 TTS; creator shows turning old posts into narrated videos — alex_verem · 2026-10-27
- Redditor crafts photorealistic 'morning bus ride' short film entirely in ComfyUI — no3us · 2026-10-27
- New blur/anti-blur slider LoRA for MOSS lets you dial background sharpness with weights — NoMouse9610 · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27
- Krea-2 Turbo + ComfyUI for product ads: prompts and results shared — comfy-king-999 · 2026-10-27
- High-res tiled Marigold V2 depth maps arrive in ComfyUI with See Through model — AntonObukhov1 · 2026-10-27