z-image/turbo-lora

texte-vers-image

TURBO

Z-Image-Turbo LoRA API by Atlas Cloud

z-image/turbo-lora

Turbo-lora

Z-Image Turbo text-to-image generation with user-supplied LoRA weights, custom sizes, and reproducible output.

Entrée

Prompt *

Loras

MAX: 3

Taille

Largeur

Hauteur

1024 × 1024 pxPlage: 256 - 1536

Graine

Num inference steps

Guidance scale

Output format

Civitai links require your API token appended: download_url&token=YOUR_TOKEN. Get it from Civitai Account Settings → API Keys.
Hugging Face links must point to a specific file (e.g. .safetensors), not a repository directory.

Sortie

Inactif

Les images générées apparaîtront ici

Configurez vos paramètres et cliquez sur exécuter pour commencer

Votre requête coûtera $0.01 par exécution. Avec $10, vous pouvez exécuter ce modèle environ 1000 fois.

Vous pouvez continuer avec :

Image-vers-Vidéo Image-vers-Image

Paramètres

Exemple de code
import requests
import time

# Step 1: Start image generation
generate_url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
    "model": "z-image/turbo-lora",  # Required. Model name
    "prompt": "A beautiful landscape with mountains and lake",  # Required. The positive prompt for the generation
    "loras": [],  # List of LoRA weights to apply. default: []
    "size": "1024*1024",  # Output image size in width*height format. (max: 1536)
    "seed": -1,  # Seed for reproducible output
    "num_inference_steps": 9,  # Number of inference steps used for generation. (min: 1)
    "guidance_scale": 1,  # Classifier-free guidance scale
    "output_format": "png",  # Image output format. options: png | jpeg | webp
    "enable_sync_mode": False,  # Wait for the result to be generated and uploaded before returning the response
    "enable_base64_output": False,  # Return base64 image data instead of an image URL
}

generate_response = requests.post(generate_url, headers=headers, json=data)
generate_result = generate_response.json()
prediction_id = generate_result["data"]["id"]

# Step 2: Poll for result
poll_url = f"https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}"

def check_status():
    while True:
        response = requests.get(poll_url, headers={"Authorization": "Bearer $ATLASCLOUD_API_KEY"})
        result = response.json()

        if result["data"]["status"] == "completed":
            print("Generated image:", result["data"]["outputs"][0])
            return result["data"]["outputs"][0]
        elif result["data"]["status"] == "failed":
            raise Exception(result["data"]["error"] or "Generation failed")
        else:
            # Still processing, wait 2 seconds
            time.sleep(2)

image_url = check_status()

Installer

Installez le package requis pour votre langage.

pip install requests

Authentification

Toutes les requêtes API nécessitent une authentification via une clé API. Vous pouvez obtenir votre clé API depuis le tableau de bord Atlas Cloud.

export ATLASCLOUD_API_KEY="your-api-key-here"

En-têtes HTTP

import os

API_KEY = os.environ.get("ATLASCLOUD_API_KEY")
headers = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {API_KEY}"
}

Protégez votre clé API

N'exposez jamais votre clé API dans du code côté client ou dans des dépôts publics. Utilisez plutôt des variables d'environnement ou un proxy backend.

Soumettre une requête

import requests

url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
    "model": "your-model",
    "prompt": "A beautiful landscape"
}

response = requests.post(url, headers=headers, json=data)
print(response.json())

Soumettre une requête

Soumettez une requête de génération asynchrone. L'API renvoie un identifiant de prédiction que vous pouvez utiliser pour vérifier le statut et récupérer le résultat.

POST/api/v1/model/generateImage

Corps de la requête

import requests

url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer $ATLASCLOUD_API_KEY"
}

data = {
    "model": "z-image/turbo-lora",
    "prompt": "A beautiful landscape with mountains and lake"
}

response = requests.post(url, headers=headers, json=data)
result = response.json()

print(f"Prediction ID: {result['data']['id']}")
print(f"Status: {result['data']['status']}")

Réponse

{
  "code": 200,
  "data": {
    "id": "pred_abc123",
    "status": "processing",
    "model": "model-name",
    "created_at": "2025-01-01T00:00:00Z"
  }
}

Vérifier le statut

Interrogez le point de terminaison de prédiction pour vérifier le statut actuel de votre requête.

GET/api/v1/model/prediction/{prediction_id}

Exemple d'interrogation

import requests
import time

prediction_id = "pred_abc123"
url = f"https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}"
headers = { "Authorization": "Bearer $ATLASCLOUD_API_KEY" }

while True:
    response = requests.get(url, headers=headers)
    result = response.json()
    status = result["data"]["status"]
    print(f"Status: {status}")

    if status in ["completed", "succeeded"]:
        output_url = result["data"]["outputs"][0]
        print(f"Output URL: {output_url}")
        break
    elif status == "failed":
        print(f"Error: {result['data'].get('error', 'Unknown')}")
        break

    time.sleep(3)

Valeurs de statut

processingLa requête est encore en cours de traitement.

completedLa génération est terminée. Les résultats sont disponibles.

succeededLa génération a réussi. Les résultats sont disponibles.

failedLa génération a échoué. Vérifiez le champ d'erreur.

Réponse terminée

{
  "data": {
    "id": "pred_abc123",
    "status": "completed",
    "outputs": [
      "https://storage.atlascloud.ai/outputs/result.png"
    ],
    "metrics": {
      "predict_time": 8.3
    },
    "created_at": "2025-01-01T00:00:00Z",
    "completed_at": "2025-01-01T00:00:10Z"
  }
}

Téléverser des fichiers

Téléversez des fichiers vers le stockage Atlas Cloud et obtenez une URL utilisable dans vos requêtes API. Utilisez multipart/form-data pour le téléversement.

POST/api/v1/model/uploadMedia

Exemple de téléversement

import requests

url = "https://api.atlascloud.ai/api/v1/model/uploadMedia"
headers = { "Authorization": "Bearer $ATLASCLOUD_API_KEY" }

with open("image.png", "rb") as f:
    files = {"file": ("image.png", f, "image/png")}
    response = requests.post(url, headers=headers, files=files)

result = response.json()
download_url = result["data"]["download_url"]
print(f"File URL: {download_url}")

Réponse

{
  "data": {
    "download_url": "https://storage.atlascloud.ai/uploads/abc123/image.png",
    "file_name": "image.png",
    "content_type": "image/png",
    "size": 1024000
  }
}

Schema d'entrée

Les paramètres suivants sont acceptés dans le corps de la requête.

Total: 10Requis: 2Optionnel: 8

Civitai links require your API token appended: download_url&token=YOUR_TOKEN. Get it from Civitai Account Settings → API Keys.
Hugging Face links must point to a specific file (e.g. .safetensors), not a repository directory.

modelstringrequired

Model name.

Default: "z-image/turbo-lora"

promptstringrequired

The positive prompt for the generation.

lorasarray[object]

List of LoRA weights to apply.

Default: []Max items: 3

pathstringrequired

URL or path to the LoRA weights.

scalenumber

Scale used before merging the LoRA weights.

Default: 1Min: 0Max: 4

sizestring

Output image size in width*height format.

Default: "1024*1024"Max: 1536

seedinteger

Seed for reproducible output. Use -1 for random.

Default: -1

num_inference_stepsinteger

Number of inference steps used for generation.

Default: 9Min: 1

guidance_scalenumber

Classifier-free guidance scale. The default value follows the Z-Image Turbo path.

Default: 1

output_formatstring

Image output format.

Default: "png"

pngjpegwebp

enable_sync_modeboolean

Wait for the result to be generated and uploaded before returning the response. This property is only available through the API.

Default: false

enable_base64_outputboolean

Return base64 image data instead of an image URL. This property is only available through the API.

Default: false

Exemple de corps de requête

{
  "model": "z-image/turbo-lora",
  "prompt": "A beautiful landscape",
  "loras": [],
  "size": "1024*1024",
  "seed": -1,
  "num_inference_steps": 9,
  "guidance_scale": 1,
  "output_format": "png",
  "enable_sync_mode": false,
  "enable_base64_output": false
}

Schema de sortie

L'API renvoie une réponse de prédiction avec les URL des résultats générés.

created_atstring

ISO timestamp when the request was created.

idstring

Unique prediction id.

modelstring

Model id used for the prediction.

outputsarray

Generated image URLs or base64 payloads.

statusstring

Task status.

Exemple de réponse

{
  "id": "pred_abc123",
  "status": "completed",
  "model": "model-name",
  "outputs": [
    "https://storage.atlascloud.ai/outputs/result.png"
  ],
  "metrics": {
    "predict_time": 8.3
  },
  "created_at": "2025-01-01T00:00:00Z",
  "completed_at": "2025-01-01T00:00:10Z"
}

Atlas Cloud Skills

Atlas Cloud Skills intègre plus de 400 modèles d'IA directement dans votre assistant de codage IA. Une seule commande pour installer, puis utilisez le langage naturel pour générer des images, des vidéos et discuter avec des LLM.

Clients pris en charge

Claude Code

OpenAI Codex

Gemini CLI

Cursor

Windsurf

VS Code

Trae

GitHub Copilot

Cline

Roo Code

Amp

Goose

Replit

40+ clients pris en charge

Installer

npx skills add AtlasCloudAI/atlas-cloud-skills

Configurer la clé API

Obtenez votre clé API depuis le tableau de bord Atlas Cloud et définissez-la comme variable d'environnement.

export ATLASCLOUD_API_KEY="your-api-key-here"

Fonctionnalités

Une fois installé, vous pouvez utiliser le langage naturel dans votre assistant IA pour accéder à tous les modèles Atlas Cloud.

Génération d'imagesGénérez des images avec des modèles comme Nano Banana 2, Z-Image, et plus encore.

Création de vidéosCréez des vidéos à partir de texte ou d'images avec Kling, Vidu, Veo, etc.

Chat LLMDiscutez avec Qwen, DeepSeek et d'autres grands modèles de langage.

Téléversement de médiasTéléversez des fichiers locaux pour l'édition d'images et les workflows image-vers-vidéo.

En savoir plus

github.com/AtlasCloudAI/atlas-cloud-skills

Serveur MCP

Le serveur MCP Atlas Cloud connecte votre IDE avec plus de 400 modèles d'IA via le Model Context Protocol. Compatible avec tout client compatible MCP.

Clients pris en charge

Cursor

VS Code

Windsurf

Claude Code

OpenAI Codex

Gemini CLI

Cline

Roo Code

100+ clients pris en charge

Installer

npx -y atlascloud-mcp

Configuration

Ajoutez la configuration suivante au fichier de paramètres MCP de votre IDE.

{
  "mcpServers": {
    "atlascloud": {
      "command": "npx",
      "args": [
        "-y",
        "atlascloud-mcp"
      ],
      "env": {
        "ATLASCLOUD_API_KEY": "your-api-key-here"
      }
    }
  }
}

Outils disponibles

atlas_generate_imageGénérez des images à partir de prompts textuels.

atlas_generate_videoCréez des vidéos à partir de texte ou d'images.

atlas_chatDiscutez avec de grands modèles de langage.

atlas_list_modelsParcourez plus de 400 modèles d'IA disponibles.

atlas_quick_generateCréation de contenu en une étape avec sélection automatique du modèle.

atlas_upload_mediaTéléversez des fichiers locaux pour les workflows API.

En savoir plus

github.com/AtlasCloudAI/mcp-server

Schéma API

{
  "components": {
    "schemas": {
      "Input": {
        "type": "object",
        "required": [
          "model",
          "prompt"
        ],
        "properties": {
          "model": {
            "type": "string",
            "description": "Model name.",
            "default": "z-image/turbo-lora"
          },
          "prompt": {
            "type": "string",
            "description": "The positive prompt for the generation."
          },
          "loras": {
            "type": "array",
            "description": "List of LoRA weights to apply.",
            "default": [],
            "maxItems": 3,
            "items": {
              "$ref": "#/components/schemas/LoraWeight"
            }
          },
          "size": {
            "type": "string",
            "description": "Output image size in width*height format.",
            "default": "1024*1024",
            "maximum": 1536
          },
          "seed": {
            "type": "integer",
            "description": "Seed for reproducible output. Use -1 for random.",
            "default": -1
          },
          "num_inference_steps": {
            "type": "integer",
            "description": "Number of inference steps used for generation.",
            "default": 9,
            "minimum": 1
          },
          "guidance_scale": {
            "type": "number",
            "description": "Classifier-free guidance scale. The default value follows the Z-Image Turbo path.",
            "default": 1
          },
          "output_format": {
            "type": "string",
            "description": "Image output format.",
            "default": "png",
            "enum": [
              "png",
              "jpeg",
              "webp"
            ]
          },
          "enable_sync_mode": {
            "type": "boolean",
            "description": "Wait for the result to be generated and uploaded before returning the response. This property is only available through the API.",
            "default": false,
            "disabled": true
          },
          "enable_base64_output": {
            "type": "boolean",
            "description": "Return base64 image data instead of an image URL. This property is only available through the API.",
            "default": false,
            "disabled": true
          }
        },
        "x-order-properties": [
          "model",
          "prompt",
          "loras",
          "size",
          "seed",
          "num_inference_steps",
          "guidance_scale",
          "output_format",
          "enable_sync_mode",
          "enable_base64_output"
        ]
      },
      "LoraWeight": {
        "type": "object",
        "required": [
          "path"
        ],
        "properties": {
          "path": {
            "type": "string",
            "description": "URL or path to the LoRA weights."
          },
          "scale": {
            "type": "number",
            "description": "Scale used before merging the LoRA weights.",
            "default": 1,
            "minimum": 0,
            "maximum": 4
          }
        },
        "x-order-properties": [
          "path",
          "scale"
        ]
      },
      "PredictionResponse": {
        "type": "object",
        "properties": {
          "created_at": {
            "type": "string",
            "format": "date-time",
            "description": "ISO timestamp when the request was created."
          },
          "has_nsfw_contents": {
            "type": "array",
            "description": "NSFW detection result for each output.",
            "items": {
              "type": "boolean"
            }
          },
          "id": {
            "type": "string",
            "description": "Unique prediction id."
          },
          "model": {
            "type": "string",
            "description": "Model id used for the prediction."
          },
          "outputs": {
            "type": "array",
            "description": "Generated image URLs or base64 payloads.",
            "items": {
              "type": "string"
            }
          },
          "status": {
            "type": "string",
            "description": "Task status."
          },
          "urls": {
            "type": "object",
            "description": "Related API endpoints."
          }
        }
      }
    },
    "securitySchemes": {
      "apiKeyAuth": {
        "type": "apiKey",
        "in": "header",
        "name": "Authorization"
      }
    }
  },
  "info": {
    "title": "AtlasCloud API",
    "version": "1.0.0",
    "description": "The AtlasCloud API."
  },
  "openapi": "3.0.0",
  "paths": {
    "/api/v1/model/prediction/{request_id}": {
      "get": {
        "parameters": [
          {
            "in": "path",
            "name": "request_id",
            "required": true,
            "schema": {
              "type": "string",
              "description": "Request ID"
            }
          }
        ],
        "responses": {
          "200": {
            "description": "Result of the request.",
            "content": {
              "application/json": {
                "schema": {
                  "$ref": "#/components/schemas/PredictionResponse"
                }
              }
            }
          }
        }
      },
      "x-api-name": "model_result"
    },
    "/api/v1/model/generateImage": {
      "post": {
        "requestBody": {
          "required": true,
          "content": {
            "application/json": {
              "schema": {
                "$ref": "#/components/schemas/Input"
              }
            }
          }
        },
        "responses": {
          "200": {
            "description": "The request status.",
            "content": {
              "application/json": {
                "schema": {
                  "$ref": "#/components/schemas/PredictionResponse"
                }
              }
            }
          }
        }
      },
      "x-api-name": "model_run"
    }
  },
  "servers": [
    {
      "url": "https://api.atlascloud.ai"
    }
  ]
}

Template de Prompt pour LLM

# z-image/turbo-lora

> Z-Image Turbo text-to-image generation with user-supplied LoRA weights, custom sizes, and reproducible output.


## Overview

- **Submit endpoint (POST)**: `https://api.atlascloud.ai/api/v1/model/generateImage` — start an async generation; returns a `prediction_id`
- **Poll endpoint (GET)**: `https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}` — poll this until the prediction finishes
- **Model ID**: `z-image/turbo-lora`


## API Information

This model can be used via our HTTP API or more conveniently via our client libraries.
See the input and output schema below, as well as the usage examples.


### Input Schema

The API accepts the following input parameters:

- **`model`** (`string`, _required_):
  Model name.
  - Default: `"z-image/turbo-lora"`

- **`prompt`** (`string`, _required_):
  The positive prompt for the generation.

- **`loras`** (`array`, _optional_):
  List of LoRA weights to apply.
  - Default: `[]`
  - Max items: 3

- **`size`** (`string`, _optional_):
  Output image size in width*height format.
  - Default: `"1024*1024"`
  - Max: 1536

- **`seed`** (`integer`, _optional_):
  Seed for reproducible output. Use -1 for random.
  - Default: `-1`

- **`num_inference_steps`** (`integer`, _optional_):
  Number of inference steps used for generation.
  - Default: `9`
  - Min: 1

- **`guidance_scale`** (`number`, _optional_):
  Classifier-free guidance scale. The default value follows the Z-Image Turbo path.
  - Default: `1`

- **`output_format`** (`string`, _optional_):
  Image output format.
  - Default: `"png"`
  - Options: "png", "jpeg", "webp"

- **`enable_sync_mode`** (`boolean`, _optional_):
  Wait for the result to be generated and uploaded before returning the response. This property is only available through the API.
  - Default: `false`

- **`enable_base64_output`** (`boolean`, _optional_):
  Return base64 image data instead of an image URL. This property is only available through the API.
  - Default: `false`



**Required Parameters Example**:

```json
{
  "model": "z-image/turbo-lora",
  "prompt": ""
}
```


**Full Example**:

```json
{
  "model": "z-image/turbo-lora",
  "prompt": "",
  "loras": [],
  "size": "1024*1024",
  "seed": -1,
  "num_inference_steps": 9,
  "guidance_scale": 1,
  "output_format": "png",
  "enable_sync_mode": false,
  "enable_base64_output": false
}
```


### Output Schema

The API returns the following output format:


- **`created_at`** (`string`, _optional_):
  ISO timestamp when the request was created.

- **`has_nsfw_contents`** (`array[boolean]`, _optional_):
  NSFW detection result for each output.

- **`id`** (`string`, _optional_):
  Unique prediction id.

- **`model`** (`string`, _optional_):
  Model id used for the prediction.

- **`outputs`** (`array[string]`, _optional_):
  Generated image URLs or base64 payloads.

- **`status`** (`string`, _optional_):
  Task status.

- **`urls`** (`object`, _optional_):
  Related API endpoints.



**Example Response**:

```json
{
  "created_at": "",
  "has_nsfw_contents": [],
  "id": "",
  "model": "",
  "outputs": [
    ""
  ],
  "status": "",
  "urls": {}
}
```


## Usage Examples

### cURL

```bash
# Step 1: Start generation (async)
curl -X POST "https://api.atlascloud.ai/api/v1/model/generateImage" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "z-image/turbo-lora",
  "prompt": "",
  "loras": [],
  "size": "1024*1024",
  "seed": -1,
  "num_inference_steps": 9,
  "guidance_scale": 1,
  "output_format": "png",
  "enable_sync_mode": false,
  "enable_base64_output": false
}'

# Response will contain: {"code": 200, "data": {"id": "prediction_id", "status": "processing"}}

# Step 2: Poll for result (replace {prediction_id} with the id returned above)
curl -X GET "https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY"

# Keep polling until status is "completed", "succeeded" or "failed"
# When completed, outputs will contain the generated content URL(s)
```

## Additional Resources

### Documentation

- [Model Playground](https://www.atlascloud.ai/models/z-image/turbo-lora)

A dramatic cinematic movie poster with dark atmospheric lighting, central silhouette of a lone figure, smoky background, bold large title text “BEYOND THE SHADOWS”, smaller tagline “Every secret has a price”, credit block at bottom, lens flare, professional Hollywood poster style

A hyper-realistic close-up of a giant ramen bowl with a tiny glowing UFO slowly descending onto the noodles, steam rising, rich broth reflections, surreal but delicious-looking detail

An ultra-detailed robot sitting in a quiet garden carefully repairing a broken flower stem with tiny mechanical tools, soft sunlight, beautiful depth of field, metallic and organic textures blending naturally

A photorealistic macro shot of a fragile soap bubble floating in the air, but inside it is a vivid swirling galaxy with stars and nebulae, high detail, cosmic glow, razor-sharp reflections on bubble surface

a gigantic neon-colored snake coiled around a disco stage, mirror-ball scales reflecting rainbow lights, people dancing below, hyper-realistic glitter particles floating through the air

Chargement...

Z-Image Turbo - Génération Texte vers Image Ultra-Rapide

NOUVEAU

Modèle de 6 Milliards de Paramètres d'Alibaba TONGYIMAI

Z-Image Turbo est le modèle texte vers image open source classé #1, surpassant FLUX.2 [dev], HunyuanImage 3.0 et Qwen-Image sur l'Artificial Analysis Image Arena. Construit par l'équipe Tongyi-MAI d'Alibaba (une division distincte de Qwen/Wan), ce modèle de 6 milliards de paramètres atteint une génération en moins d'une seconde grâce à une distillation avancée Decoupled-DMD tout en maintenant une qualité photoréaliste. Avec seulement 8 étapes d'inférence, il tient dans 16 Go de VRAM et livre des résultats professionnels optimisés pour les environnements de production critiques en vitesse.

Génération Ultra-Rapide

Seulement 8 étapes d'inférence (vs 20-50 pour les concurrents)
Génération en moins d'une seconde sur GPU H800
1.31-1.41× plus rapide que Qwen Image par étape
Tient dans 16 Go de VRAM (RTX 3060/4090)

Qualité Photoréaliste

Modèle open source classé #1 sur AI Arena
Rendu de texte bilingue (anglais et chinois)
Suivi robuste des instructions
Surpasse FLUX.1 [dev] et Qwen dans toutes les catégories

Portefeuille Stratégique de Modèles d'Alibaba

Alibaba propose trois systèmes spécialisés de génération d'images IA, chacun optimisé pour différents cas d'usage

Champion de Vitesse

Z-Image Turbo

Équipe Tongyi-MAI

Best For: Charges de travail de production critiques en vitesse

⚡ Plus rapide : 8 étapes, génération en moins d'une seconde
🏆 Modèle open source classé #1
💰 Plus rentable ($0.005/image)
🎯 Optimisé pour l'itération rapide

Roi de la Qualité

Qwen-Image

Équipe Qwen

Best For: Rendus finaux de qualité maximale

🎨 Photoréalisme et textures de peau incomparables
💡 Interactions d'éclairage supérieures
⏱️ Plus lent (20s vs 5-10s pour Z-Image)
🎯 Meilleur pour le travail de production haut de gamme

Pro de la Polyvalence

Wan 2.5/2.6

Équipe Wan

Best For: Polyvalence multimédia

🎬 Texte vers Vidéo + Image vers Vidéo
📹 Support multi-résolutions (480P-720P)
🔄 Synchronisation audiovisuelle
🎯 Génération de contenu multimodal

Key Insight: Z-Image Turbo est 1.31-1.41× plus rapide que Qwen-Image par étape, ce qui le rend idéal pour les applications nécessitant une génération rapide. Bien que Qwen-Image offre un photoréalisme légèrement meilleur pour les rendus finaux, Z-Image Turbo offre le meilleur équilibre entre vitesse et qualité pour les environnements de production.

Points Forts Techniques

Performance

Architecture S3-DiT

Adopte l'architecture Single-Stream Diffusion Transformer (S3-DiT) qui unifie le traitement de diverses entrées conditionnelles. Cette conception de 6 milliards de paramètres atteint des résultats professionnels sans la surcharge de calcul des modèles plus grands tout en maintenant une qualité de pointe.

Vitesse

Distillation Decoupled-DMD

Algorithme de distillation avancé avec mécanismes de CFG Augmentation et Distribution Matching permet une inférence en 8 étapes (vs 20-50 pour les concurrents). Atteint une génération en moins d'une seconde sur GPU H800 et fonctionne en douceur sur RTX 3060/4090 grand public avec 16 Go de VRAM.

Qualité

Performance Open Source de Premier Plan

Classé modèle open source #1 sur Artificial Analysis Image Arena, battant FLUX.2 [dev], HunyuanImage 3.0 et Qwen-Image. Excelle dans le rendu de texte bilingue (anglais et chinois), la génération photoréaliste et le suivi robuste des instructions. Publié sous licence Apache 2.0 pour usage commercial.

Parfait Pour

🎨

Création d'Art Numérique

📸

Photographie de Produits

📊

Matériel Marketing

🎬

Art Conceptuel

📱

Contenu Réseaux Sociaux

🖼️

Photographie de Stock

🎮

Actifs de Jeux

✨

Prototypage Créatif

Pourquoi Choisir Z-Image Turbo

⚡

Résultats Instantanés

Génération en moins d'une seconde avec zéro latence de démarrage à froid. Obtenez vos images immédiatement sans attendre.

💰

Rentable

Prix abordable de $0.005 par image. Développez vos projets créatifs sans faire exploser le budget.

🔌

API Prête à l'Emploi

Intégration simple de REST API. Commencez à générer des images en quelques minutes avec notre documentation complète.

Spécifications Techniques

Architecture du Modèle6 Milliards de Paramètres

Étapes d'Inférence8 NFEs (Nombre d'Évaluations de Fonction)

Vitesse de GénérationMoins d'une seconde sur H800, 5-10s sur GPU grand public

Besoins en VRAM16 Go (compatible RTX 3060/4090)

ArchitectureSingle-Stream Diffusion Transformer (S3-DiT)

Méthode de DistillationDecoupled-DMD avec CFG Augmentation

LicenceApache 2.0 (Usage Commercial Autorisé)

Classement#1 Open Source sur Artificial Analysis Arena

Prix$0.005 par Image

Commencez à Créer avec Z-Image Turbo

Découvrez la génération d'images photoréalistes ultra-rapide dès aujourd'hui. Aucune configuration requise, appelez simplement notre API et commencez à créer.

Pas de démarrage à froid - génération instantanée

Prix abordable - $0.005 par image

Résultats de qualité professionnelle

Z-Image-Turbo LoRA — 6B-parameter, ultra-fast text-to-image with custom styles

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.

Pricing

Simple per-image billing:

$0.01 per generated image

Découvrir des modèles similaires

NEW

outils d'image

Face Swap

Swap a face onto any photo. The facial identity is transferred while pose, clothing and background stay intact.

Photo Cleanup

Restore single images by reducing dust, scratches, noise, and repeated micro-texture artifacts without changing the image content.

Openai GPT Image 2 Text-to-Image

GPT Image 2 text to image is OpenAI's fast, cost-efficient text-to-image generator powered by GPT-5 guidance. Create photorealistic shots, product renders, concept art, and stylized graphics from natural-language prompts (optionally conditioned with an image). Supports custom aspect ratios, seeds, negative prompts, hex color hints, and style presets. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Openai GPT Image 2 Edit

GPT Image 2 Edit is OpenAI's image model for precise, natural-language edits. Add/remove objects, swap backgrounds, retouch faces, adjust colors/lighting, edit text/graphics, crop/resize, and apply hex color control. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2 Developer Edit

GPT Image 2 Developer Edit applies natural-language instructions to one or more reference images, with common aspect ratios and 1k, 2k, or supported 4k output tiers. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

GPT Image 2 Developer Text-to-Image

GPT Image 2 Developer Text-to-Image generates polished visuals from natural-language prompts, with common aspect ratios and 1k, 2k, or supported 4k output tiers. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Seedream v5.0 Pro Edit

ByteDance flagship next-generation image editing model. Supports up to 10 reference images while preserving identity, lighting, and color tones for professional-quality modifications.

Seedream v5.0 Pro Text-to-Image

ByteDance flagship next-generation image generation model with stronger prompt adherence, refined typography, and photorealistic detail. Single-image output at 1.5K and 2K tiers with JPEG and PNG support.

Nano Banana 2 Lite Edit Developer

Google's fastest and most cost-efficient Nano Banana image model for editing, applying natural-language edits and multi-image composition to up to 14 reference images with low latency.

Nano Banana 2 Lite Text-to-Image Developer

Google's fastest and most cost-efficient Nano Banana image model, turning natural-language text prompts into high-quality 1k images in as little as 4 seconds for rapid, high-volume generation.

Nano Banana 2 Lite Edit

Nano banana lite is the efficiency-focused model in the image generation family. Sub-2 second latency with cost-effective generation and editing, fast multi-turn local edits, and 14 supported aspect ratios.

Nano Banana 2 Lite Text-to-image

MAI-Image-2.5-Flash Text-to-image

Microsoft's fast, cost-optimized text-to-image generation model, creating high-quality images at lower cost using the same diffusion-based architecture as MAI-Image-2.5.

From

$0.03/IMAGE