Z Image Turbo API Documentation

Z Image Turbo API Documentation

Playground

Z-Image-Turbo is a 6 billion parameter text-to-image model that generates photorealistic images in sub-second time. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

Z-Image-Turbo is a 6B-parameter text-to-image model from Tongyi-MAI, engineered for production workloads where latency and throughput really matter. It uses only 8 sampling steps to render a full image, achieving sub-second latency on data-center GPUs and running comfortably on many 16 GB VRAM consumer cards.

Ultra-fast generation with production-ready quality

Where many diffusion models need dozens of steps, Z-Image-Turbo is aggressively optimised around an 8-step sampler. That keeps inference extremely fast while still delivering photorealistic images and reliable on-image text, making it a strong fit for interactive products, dashboards, and large-scale backends—not just offline batch jobs.

Why it looks so good?

• Photorealistic output at speed Generates high-fidelity, realistic images that work for product photos, hero banners, and UI visuals without multi-second waits.

• Bilingual prompts and text Understands prompts in English and Chinese, and can render multilingual text directly in the image—helpful for cross-market campaigns, posters, and screenshots.

• Low-latency, low-step design Only 8 function evaluations per image deliver extremely low latency, ideal for chatbots, configuration tools, design assistants, and any “click → image” experience.

• Friendly VRAM footprint Runs well in 16 GB VRAM environments, reducing hardware costs and making local or edge deployments more realistic.

• Scales for bulk generation Its efficiency makes large jobs—catalogues, continuous feed images, or auto-generated thumbnails—practical without blowing up compute budgets.

• Reproducible generations A controllable seed parameter lets you recreate a previous image or generate small, controlled variations for brand safety and experimentation.

How to use

• prompt – natural-language description of the scene, style, and any on-image text (English or Chinese).
• size (width / height) – choose the output resolution; supports square and rectangular images up to high resolutions (for example, 1536 × 1536).
• seed – set to -1 for random results, or use a fixed integer to make outputs reproducible.

Pricing

Simple per-image billing:

• $0.005 per generated image

Try more models and see their difference!

• Nano Banana Pro – Text-to-Image – Google’s Nano Banana Pro (Gemini 3.0 Pro Image family) delivers high-quality multi-image generation with extremely low cost per image, ideal for large-scale applications.

• Seedream V4 – Text-to-Image – ’s high-resolution text-to-image model with rich detail and diverse styles, well suited for creative illustration and commercial visuals.

• FLUX.2 [dev] – Text-to-Image – A lightweight FLUX.2-based base model hosted by WaveSpeedAI, optimised for efficient inference and LoRA-friendly training.

Paper

• Tongyi-MAI/Z-Image-Turbo

Authentication

For authentication details, please refer to the Authentication Guide.

API Endpoints

Submit Task & Query Result

set -euo pipefail

export WAVESPEED_API_KEY="your-api-key"

REQUEST_BODY=$(cat <<'JSON'
{
  "prompt": "A cinematic ocean wave at sunrise, highly detailed",
  "size": "1024*1024",
  "strength": 0.6,
  "output_format": "jpeg"
}
JSON
)

# 1. Submit the prediction once. Never retry this request automatically:
# a retry after an unclear failure can create a second, billed task.
CHECK='The task may already exist. Check History or the predictions list before submitting again.'
if ! SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  --connect-timeout 10 --max-time 60 \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/z-image/turbo" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}"); then
  printf 'Submit failed. %s\n' "${CHECK}" >&2
  exit 1
fi

TASK=$(printf '%s' "${SUBMIT_RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
PREDICTION_ID=$(printf '%s' "${TASK}" | jq -r '.id // empty')
if [ -z "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id. %s\n' "${CHECK}" >&2
  exit 1
fi
RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"

# 2. Poll until the prediction reaches a final status.
BODY_FILE=$(mktemp)
trap 'rm -f "${BODY_FILE}"' EXIT
while true; do
  # Transient errors (429, 5xx) are retried, honoring Retry-After.
  HTTP_CODE=$(curl --silent --show-error --fail-with-body \
    --connect-timeout 10 --max-time 60 --retry 5 --retry-delay 2 \
    --output "${BODY_FILE}" --write-out '%{http_code}' \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}" || true)
  case "${HTTP_CODE}" in
    2??) ;;
    000|429|5??)  # dropped connection or temporary error: keep polling
      printf 'Task %s: HTTP %s, polling again in 10s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      sleep 10; continue ;;
    *) printf 'Task %s: HTTP %s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      cat "${BODY_FILE}" >&2; exit 1 ;;
  esac
  RESULT=$(jq 'if type == "object" and has("data") then .data else . end' "${BODY_FILE}")
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout|deleted)
      printf 'Task %s ended with status %s\n' "${PREDICTION_ID}" "${STATUS}" >&2
      printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    *) sleep 2 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Every model also accepts the global request options enable_sync_mode and enable_base64_output; see their pages for details.

Response Parameters

Result Request Parameters

Result Response Parameters