Image Generation with Text Prompts

Our API endpoints enable media creation with BFL models. It follows an asynchronous design, where you first make a request for a generation and then query for the result of your request.

​

API Endpoints

​

Primary Global Endpoint

api.bfl.ai - Recommended for most use cases

• Routes requests across all available clusters globally
• Automatic failover between clusters for enhanced uptime
• Intelligent load distribution prevents bottlenecks during high traffic
• Does not support finetuning and finetuned inference
• Important: Always use the polling_url returned in responses when using this endpoint

​

Regional Endpoints

api.eu.bfl.ai - European Multi-cluster

• Multi-cluster routing limited to EU regions
• GDPR compliant
• Does not support finetuning and finetuned inference

api.us.bfl.ai - US Multi-cluster

• Multi-cluster routing limited to US regions
• Does not support finetuning and finetuned inference

​

Legacy Regional Endpoints

api.eu1.bfl.ai - EU Single-cluster

• Single cluster, no automatic failover
• Required for finetuning and finetuned inference operations in EU region

api.us1.bfl.ai - US Single-cluster

• Single cluster, no automatic failover
• Required for finetuning and finetuned inference operations in US region

​

Available Endpoints

We currently support the following endpoints for image generation:

1. /flux-kontext-pro
2. /flux-kontext-max
3. /flux-pro-1.1-ultra
4. /flux-pro-1.1
5. /flux-pro
6. /flux-dev

​

Create Your First Image

​

Submit Generation Request

To submit an image generation task with FLUX 1.1 [pro], create a request:

# Install curl and jq, then run:
# Make sure to set your API key: export BFL_API_KEY="your_key_here"

request=$(curl -X 'POST' \
  'https://api.bfl.ai/v1/flux-kontext-pro' \
  -H 'accept: application/json' \
  -H "x-key: ${BFL_API_KEY}" \
  -H 'Content-Type: application/json' \
  -d '{
  "prompt": "A cat on its back legs running like a human is holding a big silver fish with its arms. The cat is running away from the shop owner and has a panicked look on his face. The scene is situated in a crowded market.",
  "aspect_ratio": "1:1",
}')

echo $request
request_id=$(jq -r .id <<< $request)
polling_url=$(jq -r .polling_url <<< $request)
echo "Request ID: ${request_id}"
echo "Polling URL: ${polling_url}"

A successful response will be a json object containing the request’s id and a polling_url that should be used to retrieve the result.

​

Poll for Results

To retrieve the result, poll the endpoint using the polling_url:

# This assumes that the request_id and polling_url variables are set from the previous step

while true
do
  sleep 0.5
  result=$(curl -s -X 'GET' \
    "${polling_url}" \
    -H 'accept: application/json' \
    -H "x-key: ${BFL_API_KEY}")
  
  status=$(jq -r .status <<< $result)
  echo "Status: $status"
  
  if [ "$status" == "Ready" ]
  then
    echo "Result: $(jq -r .result.sample <<< $result)"
    break
  elif [ "$status" == "Error" ] || [ "$status" == "Failed" ]
  then
    echo "Generation failed: $result"
    break
  fi
done

A successful response will be a JSON object containing the result, where result['sample'] is a signed URL for retrieval.

See our reference documentation for a full list of options and our inference repo.

​

Limits