Nine model pages, two families, one table. Gemini gives you more resolutions and aspect ratios; OpenAI gives you steadier layouts and type. Credits below are measured from real provider cost — no rounded guesses.
Fastest, cheapest, most detailed, best at text — find the row that matches the job instead of guessing.
More resolutions and ratios; 2 Lite is fastest, Pro is the detail tier.
| Model | Best for | Ratios | Output | Credits | Speed |
|---|---|---|---|---|---|
| Nano Banana 2 Lite | Fastest drafts and volume exploration | 14 ratios | 1K only | 4 | ~3s (fastest) |
| Nano Banana 2.1 | Product recontextualisation, newest tier | 14 ratios | 1K / 2K / 4K | 4 / 6 / 12 | ~11s |
| Nano Banana 2 (Gemini 3.1 Flash Image) | Balanced drafts and final scenes | 1:1 / 3:4 / 4:3 | 1K / 2K / 4K | 6 / 10 / 12 | ~10s |
| Nano Banana Pro (Gemini 3 Pro Image) | In-image text and material detail | 1:1 / 3:4 / 4:3 | 1K / 2K / 4K | 14 / 24 | slowest |
Stronger layouts and type; Flare favours speed, Sunburst precise edits.
| Model | Best for | Ratios | Output | Credits | Speed |
|---|---|---|---|---|---|
| GPT Image 2.5 Flare | Cheapest OpenAI exploration | 1:1 / 3:4 / 16:9 | 1K | 1 / 2 / 6 | fast |
| GPT Image 2.5 Sunburst | Precise edits, keep the layout | 1:1 / 3:4 / 16:9 | 1K | 1 / 2 / 6 | fast |
| GPT Image 2.6 Flare | Presentable first pass at speed | 1:1 / 3:4 / 16:9 | 1K | 3 / 5 / 6 | fast |
| GPT Image 2.6 Sunburst | Dense texture and exact edits | 1:1 / 3:4 / 16:9 | 1K | 3 / 5 / 6 | fast |
| GPT Image 2 | Text-rich layouts and prompt-faithful structure | 1:1 / 3:2 / 2:3 | 1K | 2–25 (quality tiers) | medium |
Credits are per image and measured from live API cost; failed generations are refunded. Gemini 2 Lite is the fastest, 2.1 the newest Gemini tier, GPT Image 2.5 Flare the cheapest OpenAI tier, and Gemini Pro the detail tier.
How the models differ, and which one to start with.
Open the editor, pick a model and generate a first frame — switch tiers whenever the brief changes.