To generate an AI image on Oakgen, open the image generator, pick a model, write a prompt, set the size and quality, and press generate. The cost appears before you run it. Oakgen's eight text-to-image models cost between 2 and 102 credits per image, according to its pricing code on October 25, 2026. You need a paid plan, which started at $19 a month.
This guide was rewritten in October 2026. The first version recommended models Oakgen no longer runs, such as Flux 2 Pro, Ideogram V3, Recraft V3 and Seedream.
What do you need before you start?
A paid Oakgen plan and an idea you can describe in a sentence or two. Oakgen has no free credits or free trial. On the pricing page on October 30, 2026, plans were:
| Plan | Monthly | Billed yearly | Credits a month |
|---|---|---|---|
| Basic | $19 | $19 a month | 4,000 |
| Pro | $39 | $35 a month | 8,000 |
| Ultimate | $69 | $65 a month | 15,000 |
| Creator | $99 | $85 a month | 30,000 |
Plan prices can change. Monthly plan credits reset when your plan renews and do not roll over. Credits you buy as add-ons never expire and are used after your plan credits. Yearly plans hand out credits monthly.
Which image model should you choose?
Pick by the job and the budget. These are the eight text-to-image models on Oakgen as of October 25, 2026, with prices from Oakgen's pricing code.
| Model | Good first choice for | Credits per image | Options worth knowing |
|---|---|---|---|
| Grok Imagine Image | Cheap drafts and idea testing | 7 | 10 aspect ratios, up to 4 images per run |
| Nano Banana 2.1 | Low-cost everyday images | 14 at 1K, 46 at 4K | Resolution up to 4K |
| Nano Banana 2 | All-round images in any shape | 21 at 512, 28 at 1K, 55 at 4K | 10 aspect ratios, seed, web search |
| Nano Banana Pro | Detailed images at 4K | 51 at 1K or 2K, 102 at 4K | Auto aspect ratio, seed, web search |
| GPT Image 2 | Following long, detailed briefs | 2 low, 14 to 18 medium, 56 to 72 high | Prompts up to 32,000 characters |
| GPT Image 2.5 Flare | Text in images, faster runs | 5 low, 7 medium, 21 high at 1024x1024 | Transparent background, five quality levels |
| GPT Image 2.5 Sunburst | Text in images, hardest briefs | 5 low, 7 medium, 21 high at 1024x1024 | Transparent background, five quality levels |
Oakgen's image generator page describes the GPT Image models as prioritising prompt fidelity and in-image text. OpenAI calls Sunburst its most capable image model and Flare its fastest. We did not run a quality test for this guide, so treat the "good first choice" column as a starting point and compare models on your own prompt in Image Arena.
How do you generate your first image?
- Open the image generator and start creating. Signed in, this opens Create with Image selected.
- Pick a model from the model picker. Nano Banana 2 is a safe first choice.
- Type your prompt in the prompt box.
- Open Settings and set the aspect ratio or size, resolution or quality, and the number of images, from one to four.
- Check the credit cost, then generate.
- Download the image you like, or use it as a reference for the next run.
If a generation fails, the credits come back automatically and you can retry.
How do you write a prompt that works?
Lead with the subject, then add setting, style, light and camera, in that order. One to three clear sentences beat a long list of conflicting instructions.
Instead of "a cat", try:
"A fluffy orange tabby cat on a windowsill, afternoon sun casting warm shadows, plants in the background, pet photography, shallow depth of field, warm colours"
What to include:
- Subject. What it is, what it looks like, what it is doing.
- Setting. Indoors, outdoors, a plain studio background.
- Style. Photo, illustration, oil painting, 3D render.
- Light. Golden hour, soft studio light, neon.
- Camera. 85mm portrait, wide angle, overhead, macro.
- Exact text. If the image needs words, put them in quotes, for example a poster that reads "OPEN LATE".
- "A medieval castle on a cliff at sunset, dramatic clouds, oil painting style, warm colours, detailed stonework"
- "Studio headshot of a confident businesswoman, soft key light, neutral grey background, 85mm lens, shallow depth of field"
- "Isometric 3D render of a small coffee shop interior, warm lighting, miniature style, pastel colours"
None of the current models has a negative prompt field. Say what you want ("plain white background") rather than what you do not want.
Which settings matter most?
Four settings change the result and the price.
- Aspect ratio or size. Set it to the final use before you generate: 1:1 for feed posts, 16:9 for thumbnails and slides, 9:16 for stories and reels, 21:9 for banners. Cropping later loses parts of the picture.
- Resolution or quality. On Nano Banana models it is 512, 1K, 2K or 4K. On GPT Image models it is low, medium or high (GPT Image 2.5 adds xhigh and max). Higher costs more, so draft low and rerun the keeper high.
- Number of images. One to four per run. Each image is charged.
- Seed. Nano Banana 2 and Nano Banana Pro accept a seed. Reuse it with a small prompt change to keep a similar composition.
How do you edit an image or use references?
Switch to an edit model and upload your image. Nano Banana 2 Edit takes up to 10 input images, which helps keep a character or product the same across a set. GPT Image 2 Edit takes up to 4 reference images and an optional mask to limit the change to one area. Most edit models cost the same as their text-to-image versions.
For other jobs, use the dedicated tools.
- Image Editor to edit, retouch and restyle a photo
- Image Upscaler, where Topaz Upscale costs 28 credits at 2x and 109 at 4x
- Photo Studio for product and portrait shots
- Avatar Generator for a consistent character
What are the most common beginner mistakes?
- Too vague. "A landscape" gives you something generic. Name the place, time of day, weather and viewpoint.
- Wrong shape. A square image will not crop into a good 16:9 banner. Set the ratio first.
- One model only. The same prompt reads differently on each model. Spend a few credits on two or three before a big batch.
- Final quality on every draft. GPT Image 2 at high quality is 56 to 72 credits. At low quality it is 2.
- Overloaded prompts. Two hundred words of competing instructions confuse the model. Keep the important parts.
What can you do with the image next?
- Animate it in the AI video generator, using it as the start frame for an image-to-video model such as Kling v3 Pro.
- Make a portrait talk with Talking Photo.
- Turn an object image into a 3D model with Image to 3D, which costs 271 credits per model.
- Add a soundtrack with the Music Generator.
For model-by-model picks, see best AI image generators by use case.
How we checked this
We read the eight text-to-image model configs in Oakgen's image gateway, the edit model configs and Oakgen's pricing code on October 25, 2026, and plan prices on the pricing page on October 30, 2026. The OpenAI descriptions come from OpenAI's model pages, read the same day. We did not run a quality comparison between models for this guide.
Make your first image
Pick a model, write a prompt and see the credit cost before you generate.
Frequently asked questions
Which image model should a beginner start with on Oakgen?
Start with Nano Banana 2 at 1K, which costs 28 credits an image and has the widest choice of aspect ratios, or Grok Imagine Image at 7 credits for cheap drafts. Move to GPT Image 2.5 when the image needs readable text. Prices are from Oakgen's pricing code on October 25, 2026.
How many credits does one AI image cost on Oakgen?
Between 2 and 102 credits, depending on the model and settings, according to Oakgen's pricing code on October 25, 2026. GPT Image 2 at low quality is the cheapest at 2 credits, and Nano Banana Pro at 4K is the most expensive at 102. The generator shows the cost before you run it, and failed generations are refunded.
Can I generate AI images on Oakgen for free?
No. Oakgen has no free credits, free tier or free trial, so you need a paid plan. On the pricing page on October 30, 2026, plans started at $19 a month for 4,000 credits. Monthly plan credits reset when the plan renews and do not roll over. Add-on credits never expire and are used after plan credits.
Do Oakgen image models support negative prompts?
No. None of the eight text-to-image models on Oakgen has a negative prompt field as of October 25, 2026. Describe what you want instead, for example 'clean background' rather than 'no clutter'.
Can I edit an image I already have?
Yes. Pick an edit model and upload your image as a reference. Nano Banana 2 Edit accepts up to 10 input images and GPT Image 2 Edit up to 4, plus an optional mask. For manual touch-ups, the Image Editor and Image Upscaler are separate tools.
Can I use the images commercially?
Oakgen does not claim ownership of eligible outputs. Under section 4.4 of Oakgen's Terms, your right to use an image depends on following the Terms, the law, the model provider's terms and the rights you hold in anything you upload, such as a logo, a product design or a person's likeness.

