Poe API

Qwen-Image-2.0

Qwen Image 2.0 is Alibaba's unified image generation and editing model, excelling at complex text rendering in both Chinese and English, realistic textures, precise semantic adherence, and multi-image fusion. The Pro variant delivers enhanced detail fidelity and professional-grade text-in-image accuracy, while the Standard variant balances quality with faster response times. Notes: - This model is served from the Singapore region. - Upload 1-3 images to enable image editing mode. Without images, the bot operates in text-to-image generation mode. - Responses may take 10-30 seconds depending on resolution, number of images, and prompt complexity. - Positive prompt has a max of 800 characters, negative prompt has a max of 500 characters. Parameter controls available: 1. Generation - Model: Select from Qwen Image 2.0 Pro or Qwen Image 2.0. Pro offers enhanced text rendering, realism, and semantic adherence. Standard is faster and lower cost. (Default: Pro) - Image Size: Select from (16:9, 9:16, 1:1, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5). Mapped resolutions: 16:9 = 1920x1080, 3:2 = 1536x1024, 4:3 = 1440x1080, 5:4 = 1280x1024, 1:1 = 1024x1024, 4:5 = 1024x1280, 3:4 = 1080x1440, 2:3 = 1024x1536, 9:16 = 1080x1920. If selected, this size is sent. If not selected: text-to-image defaults to 16:9; image editing uses model defaults based on input image ratio (for multiple input images, the last image determines the ratio). - Custom Width / Custom Height: Shown when Custom Size is enabled. Enter pixel values between 512 and 2048. Total pixels must be between 512x512 and 2048x2048. (Default: disabled) - Number of Images: Generate 1 to 6 images per request. Each image is billed separately. (Default: 1) 2. Advanced - Prompt Enhancement: Automatically optimizes short prompts using AI for better results. Adds 3-5 seconds of latency. (Default: enabled) - Negative Prompt: Describe content to exclude from the generated image (e.g. "low quality, blurry, distorted, extra fingers"). - Seed: Set from 0-2147483647 for reproducible results. Leave empty for random. - Watermark: Add a Qwen-Image watermark to the bottom-right corner. (Default: disabled) Attachments (Image Editing Mode): - Attach 1-3 images (JPEG, PNG, BMP, WEBP, TIFF, GIF) with a text instruction to edit them. - For best results, input image resolution should be between 384 and 3072 pixels per side. - Maximum file size per image: 10 MB. - Multi-image editing: Reference images by order (e.g. "The girl in Image 1 wears the dress from Image 2 and sits in the pose from Image 3") (images should be referenced in the order they are uploaded by). - Supported editing operations: style transfer, text modification, object addition/removal, pose changes, background replacement, viewpoint transformation, sketch-to-image, depth map generation, and more.

OFFICIAL

Build with Qwen-Image-2.0 using the Poe API

Start by creating an API key, for use with any bot on Poe:

Generate API key

Code samples

python

import openai

client = openai.OpenAI(
    api_key = "YOUR_POE_API_KEY",  # or os.getenv("POE_API_KEY")
    base_url = "https://api.poe.com/v1",
)

chat = client.chat.completions.create(
    model = "qwen-image-2.0",
    messages = [{
      "role": "user",
      "content": "Sleeping cat in cinematic film still, shallow depth of field, vignette, highly detailed"
    }]
)

print(chat.choices[0].message.content)

See the full documentation for comprehensive guidance on getting started.