Nodes/ERPK Collection/Qwen Image 2.0 Text-to-Image
ComfyUI Node

Qwen Image 2.0 Text-to-Image

The current-gen Qwen, Standard or Pro

By eRepublik-Labs·Created 11 months ago·Updated 21 days ago· 1
Qwen Image 2.0 Text-to-Image
  • client
  • image
modelQwen Image 2.0
prompt
width1024
height1024
seed-1
output_formatjpeg
enable_sync_modefalse
enable_base64_outputfalse

Qwen Image 2.0 Text-to-Image is the current-generation Qwen text-to-image node, and the README is blunt about what it is: a full replacement for the Qwen 1.x line. If you've been using the older Qwen Image nodes in this pack, this is the upgrade - same bilingual (Chinese/English) prompt handling, same style of interface, better model. It comes in two tiers: Qwen Image 2.0 (standard) and Qwen Image 2.0 Pro (higher quality, pricier).

The lineage is worth knowing: Qwen is Alibaba's, from the same lab that's been shipping the open-weights Qwen-Image the local community trains LoRAs on. The API-served 2.0 here is the commercial iteration of that family - you're paying per image, but you're getting the current-gen model without downloading anything or touching a GPU. Standard is the default for a reason; Pro is for when the render matters and the extra quality is visible.

How it works

The node sends your prompt (plus a size of width*height, output format, and sync-mode flag) to WaveSpeed's Qwen Image 2.0 endpoint, waits for the render, and converts the returned URL into a real ComfyUI IMAGE tensor. Resolution runs 256 to 2048px - notably higher than the 1536px ceiling on the older Qwen nodes, so you get more headroom for detailed scenes. The output drops straight into your image chain like any other IMAGE.

The inputs that matter

  • model - Qwen Image 2.0 or Qwen Image 2.0 Pro.
  • prompt - the description, Chinese or English.
  • width / height - 256 to 2048, step 8, default 1024×1024.
  • output_format - jpeg (default), png, or webp.
  • enable_sync_mode - wait for completion before returning (off by default, polls in background).
  • enable_base64_output - base64 instead of URL, for scripting.

The seed widget is cache control only. The optional client input wires to a WaveSpeed API Client node; unconnected, the key resolves from Settings.

Installing it

Part of the ERPK Collection (eRepublik-Labs/comfyui-nodes-erpk). ComfyUI Manager: search erpkERPK Custom Nodes → restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/eRepublik-Labs/comfyui-nodes-erpk.git erpk
cd erpk
pip install -r requirements.txt

WaveSpeed API key required (Settings → ERPK → API Keys). No model downloads - the WaveSpeed dependency is just pydantic/requests/Pillow.

Common issues

  • "Prompt is required" - blank prompts are rejected up front.
  • Pro tier costs more - the dropdown hides the price. If the bill surprises you, check which tier you're on.
  • Higher res isn't always better - 2048px renders cost more and take longer; use it for finals, not drafts.
  • Billed per image - seed on randomize means a fresh charge per queue run. Pin it for plumbing tests.
CategoryERPK/WaveSpeedAI

Inputs (9)

NameTypeDefaultDescription
modelCOMBOQwen Image 2.0Model variant: Qwen Image 2.0 (standard) or Qwen Image 2.0 Pro (higher quality)
promptSTRINGText description of the image to generate (Chinese or English)
clientoptWAVESPEED_AI_API_CLIENTWaveSpeed API client (optional if API key is configured in Settings)
widthoptINT1024256–2048Image width (256 to 2048)
heightoptINT1024256–2048Image height (256 to 2048)
seedoptINT-1-1–2147483647Random seed for reproducibility (-1 for random)
output_formatoptCOMBOjpegOutput image format
enable_sync_modeoptBOOLEANfalseWait for completion before returning response
enable_base64_outputoptBOOLEANfalseReturn BASE64-encoded output instead of URL

Outputs (1)

NameTypeDescription
imageIMAGE