With FlowModels, you can generate images from a text description or edit an existing template according to your prompt. Unlike a video, an image is created within the request: you send the prompt and get the finished image back in the same response, usually after a few seconds. You can do this in the Playground in the dashboard or directly through the API.
The models
| Model | Formats | Quality | Editing |
|---|---|---|---|
openai/gpt-image-2.6-sunburst |
square, portrait, landscape, 2k |
low, medium |
yes |
google/gemini-3.1-flash-image |
square, portrait, landscape, 2k |
medium |
no |
google/gemini-nano-banana-2.1 |
square, portrait, landscape, 2k |
medium |
no |
minimax/h3-image |
square, portrait, landscape, 2k |
medium |
yes |
minimax/h3-image-highspeed |
square, portrait, landscape, 2k |
medium |
yes |
square is square, portrait upright, landscape wide and 2k a wide landscape format in higher resolution. Without a value, square applies. Only use models that appear with the kind image in your response from GET /v1/models. There, image lists the formats, qualities and whether the model can edit, and pricing.variants the prices per image. There's no automatic switch to another model.
The API still accepts the former ID google/gemini-3.1-flash-image-preview. Google has shut that preview model down, so such a request is generated with google/gemini-3.1-flash-image and billed at its price, and the response names that model. Best switch your requests to the new ID.
In the Playground
Open the Playground in the dashboard and switch to the Images tab. Choose a Model, describe the image in the Prompt field and set Format and Quality. With models that can edit (see the table), you can also upload an Image to edit (optional) (JPEG, PNG or WebP, up to 10 MB). With an image, it is edited according to your prompt; without one, a new image is created.
Before you start, Price per image shows what your selection costs. Generate image starts the generation.

Your images shows your team's images from the last seven days, including the ones created through the API. You can open them in full size, download or delete them. The gallery doesn't show prompts because the whole team can see it. Billing runs through your team's FlowModels credit, so you don't need an API key for this. In Usage and Requests, your calls appear as "Playground · your name".
What the other Playground tabs can do and which limits apply there is explained in Try models in the Playground.
Via the API
You create a new image with POST /v1/images/generations:
curl https://api.flowmodels.one/v1/images/generations \
-H "Authorization: Bearer at_your_key" \
-H "Content-Type: application/json" \
-d '{"model": "openai/gpt-image-2.6-sunburst", "prompt": "A paper boat on a calm lake at sunrise", "size": "landscape", "quality": "low"}'
The most important fields:
modelandprompt: The prompt can be up to 4,000 characters long.sizeandquality: format and quality within the range of the model. Without a value,squareapplies and, for the OpenAI model,low.n: number of images, currently only1.response_format:url(default) returns a download address,b64_jsonthe image itself as Base64.
To edit, send the same fields to POST /v1/images/edits together with the image, either as image_url (https address or data: URL) or as the image_file_id of a file from POST /v1/files. JPEG, PNG and WebP up to 10 MiB are allowed. Currently only openai/gpt-image-2.6-sunburst can edit.
If the model doesn't know a field, the API rejects it with 422 and the field name in param before anything is reserved. With an Idempotency-Key, you can safely repeat the request after a network error, and no second image is created.
Response and retrieval
The response follows the structure of the OpenAI images API: created and data with exactly one entry. It also contains id, status, content_type, size_bytes, expires_at and usage. status is completed when the image is ready, failed after an error and in_progress while generation is still running.
The address in data[0].url points to GET /v1/images/{id}/content. This call needs your key and responds with a 302 to a download address that is valid for 15 minutes:
curl -L https://api.flowmodels.one/v1/images/{id}/content \
-H "Authorization: Bearer at_your_key" \
-o image.png
With GET /v1/images/{id}, you can query an image again later. There are no webhooks for images because the response already contains the finished image.
Retention
A finished image stays available for seven days, and expires_at gives the exact time. After that, we delete it automatically. With DELETE /v1/images/{id} or in the Playground, you delete it earlier. Store images you need for longer on your side.
What an image costs
Images are billed per image. The price depends on the model, the format and the quality. You can see the current price under Models and in GET /v1/models under pricing.variants.
Before generation, we reserve the price of one image. The amount is only charged once the image is stored with us. Rejected and failed requests cost nothing. What was billed is shown in the response under usage.cost.
Limits
- A generation takes one of your concurrent request slots, see Request limit: how the tiers work. Queries, downloads and deletions don't count.
- Each request creates exactly one image.
- The generated images are AI generated. Label them accordingly when you publish them.
If something goes wrong
422 invalid_image_request: A field doesn't fit the model, andparamnames it.422 image_input_invalid: The image to edit is missing, invalid, too large or doesn't belong to your team.422 image_request_rejected: The model rejected the request. Adjust the prompt or settings.422 image_content_rejected: The image was rejected because of its content.429 insufficient_quotaor429 spend_limit_exceeded: Your credit or spending limit isn't enough.503 image_model_unavailable,503 image_pricing_unavailableor503 image_provider_unavailable: Try again later, your credit wasn't charged.404 image_not_found: The ID is unknown, expired or doesn't belong to your team.409 image_not_readyor409 image_not_deletable: The current status doesn't allow the action.