Text to Image AI Generator & Prompt Playground
Turn a written prompt into images and compare every model in one workspace, from Flux 2, Qwen, and Nano Banana to Midjourney, Imagen 4, Seedream, and GPT-4o.
How text to image works on SeedVR2
Text to image turns a written prompt into a new picture. The model selector holds provider and open-weight models side by side, so the same prompt can be tried at different price points and compared for style, detail, and composition before you commit to a final image. Nothing has to be uploaded and nothing is installed locally, which keeps the first draft cheap enough to treat wording and model choice as the experiment itself.
From prompt to finished image
Every run follows the same path: your prompt is checked, one model renders it, and the result stays in the workspace so you can compare, download, or edit it. Each finished image keeps the prompt, model, and settings that produced it, so a frame you like can be re-run on a cheaper model for a quick variation or on a stronger one for a final export without retyping anything. Results stay in the workspace until you remove them, and the same prompt can be routed to image to image for a targeted edit instead of a full regeneration.
1. Write the prompt
Describe the subject, framing, lighting, and style you want. A prompt that names the subject, the shot type, and the mood gives the model enough to work with, while a one-line prompt leaves the composition to chance. Reuse the example inside the prompt box whenever you are unsure how much detail to add.
2. Screening and validation
Before a render starts, the request is checked for format, size, and content. Prompts or reference images that break the acceptable-use policy are rejected before any credits are spent, and the workspace reports which rule blocked the run so you can rewrite the wording and submit it again.
3. Pick a model and see the price
The selector lists every available model with its credit price, and the same price is shown again in the workspace before you press generate. Flux 2, Qwen, Nano Banana, GPT-4o, Imagen 4, Seedream, Wan 2.5, and Midjourney are all reachable from one prompt box, so a cheap draft and a premium render can be compared from the same idea.
4. What to upload
Uploads are optional on this route. Add up to five PNG, JPG, or WebP images to guide a variation or an edit, or leave the upload area empty for a pure text prompt. Files are limited to about 10 MB each, and any file the workspace cannot read is reported back instead of being silently dropped.
5. How long a run takes
Provider time depends on the model and the settings you choose. In the current catalog a default run is typically reported between 5 and 18 seconds, with the lightest models finishing first. A generation continues in the background if you close the tab, and the finished image appears in the workspace when it is ready.
6. When a different route fits better
Use text to image when no source picture exists yet. If you already have a photo to retouch or a frame to refine, image to image edits that exact picture instead of generating a new one, and the upscaler is the right tool when a real photo only needs to be enlarged or repaired.
Open image to imageWhat text to image costs
Every model in the selector is billed per image at its default entry settings, from 20 to 244 credits. Higher resolution, extra candidates, or a stronger model move the real charge, and the exact price of the model you pick is shown in the workspace before the run starts.
| Model | Credits per image |
|---|---|
| Nano Banana | 30 |
| Nano Banana Pro | 240 |
| Flux 2 Pro | 49 |
| Flux 2 Flex | 98 |
| Flux 2 Dev | 20 |
| Qwen | 50 |
| Midjourney | 244 |
| Imagen 4 | 50 |
| Seedream | 40 |
| Wan 2.5 · Text To Image | 45 |
| GPT-4o | 40 |
Text to image FAQ
Short answers about prompts, uploads, credits, and failed runs.
How are credits charged for a text to image run?
A run bills the selected model price once, taken from the table above, at the moment you start it. That price reflects the model default settings; a higher resolution or a different number of outputs can change it, and a run that never starts is never charged.
What happens if I run out of credits?
The run does not start, so nothing is charged. Add a credit pack or a subscription on the pricing page and return to the same prompt; your text and settings stay in the workspace while you top up.
Why was my prompt or image rejected?
Requests are screened against the acceptable-use policy before a model runs. If a prompt or reference image is blocked, rewrite the wording or replace the image and submit again; a blocked request is not charged because the render never began.
What if the provider times out or the render fails?
Confirmed system or provider failures are covered by the refund policy, which explains how a job is reviewed and repaid. You can also retry the same prompt, or drop to a lighter model at a lower resolution to reduce the chance of a repeat failure.
Do I need to upload a reference image?
No. Text to image works from the prompt alone. Uploads are optional and only influence variations or edits; up to five images can be attached to a single run.
Can I iterate on an image I like?
Yes. Keep the prompt and change one variable at a time, such as the model or the aspect ratio, so you can tell which change made the difference. When a frame is close, send it to image to image for a targeted edit rather than regenerating it.