Z Image Turbo LoRA

Z Image Turbo LoRA

Playground

Z-Image-Turbo LoRA (6B) enables ultra-fast text-to-image generation with external LoRA support. Generate photorealistic images in sub-second latency while applying up to 3 LoRAs for custom styles. Ready-to-use REST API, best performance, no coldstarts, affordable pricing.

Features

Z-Image-Turbo LoRA is a personalised version of Tongyi-MAI’s 6B-parameter Z-Image-Turbo model. It keeps the same 8-step, ultra-fast sampler and low VRAM footprint, while letting you plug in up to three LoRA adapters to inject your own styles, characters, or brand identity into each generation.

Ultra-fast generation with LoRA personalisation

Where many diffusion models need dozens of steps, Z-Image-Turbo LoRA stays aggressively optimised around 8 sampling steps. On top of that, it adds LoRA hooks so you can steer the visual style without retraining the base model—perfect for interactive products, dashboards, and large-scale backends that still need a branded look.

Why it looks so good

• Photorealistic output at speed Generates high-fidelity, realistic images suitable for product photos, hero banners, and UI visuals—now with your own LoRA styles layered on top.

• Bilingual prompts and text Understands prompts in English and Chinese, and can render multilingual on-image text, ideal for cross-market campaigns and UI screenshots.

• LoRA-powered customisation Attach up to 3 LoRAs per request to add a specific art style, character look, or brand aesthetics without touching the base weights.

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

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

• Scales for bulk generation The efficient sampler keeps large jobs—catalogues, continuous feeds, or mass thumbnail generation—practical, even when every image uses one or more LoRAs.

• Reproducible generations A controllable seed parameter lets you recreate previous images 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 that fits your use case.

• seed – set to -1 for random results, or use a fixed integer to make outputs reproducible.

• loras – optional list of up to three LoRA adapters:

• path – a LoRA identifier such as <owner>/<model-name> or a direct .safetensors URL.

• scale – numeric strength for that LoRA; higher values apply a stronger stylistic effect.

You can click “Add Item” in the loras panel to add 1–3 LoRAs. They are combined during generation, so a single prompt can mix, for example, a character LoRA, a style LoRA, and a brand-colour LoRA.

For detailed, step-by-step guidance on finding, uploading, and using LoRAs on WaveSpeedAI, see our LoRA tutorials How to use LoRA.

Pricing

Simple per-image billing:

• $0.01 per generated image

Try more models and compare

• stability-ai/sdxl-lora – Stability AI’s SDXL LoRA hub, offering a wide range of ready-made styles and subjects for fast, lightweight customisation on top of the SDXL base model.

• wavespeed-ai/qwen-image/edit-plus-lora – Qwen Image Edit Plus with LoRA support, combining strong semantic understanding with style-controllable, localised image editing.

• wavespeed-ai/flux-2-dev/edit-lora – FLUX.2 [dev] Edit enhanced with LoRA adapters, enabling prompt-based image editing that can also match specific art styles, characters, or brand looks.

Reference

• How to train your LoRA?
• How to use your LoRA?

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",
  "seed": -1,
  "output_format": "jpeg"
}
JSON
)

# 1. Submit the prediction.
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/z-image/turbo-lora" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}")

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 "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id
' >&2
  exit 1
fi
RESULT_URL=$(printf '%s' "${TASK}" | jq -r '.urls.get // empty')
if [ -z "${RESULT_URL}" ]; then RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"; fi

# 2. Poll until the prediction finishes.
while true; do
  RESPONSE=$(curl --silent --show-error --fail-with-body \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}")
  RESULT=$(printf '%s' "${RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout) printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    created|processing) sleep 2 ;;
    *) printf 'Unexpected status: %s
' "${STATUS}" >&2; exit 1 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Response Parameters

Result Request Parameters

Result Response Parameters