← Docs Home · Model Catalog · HTTP API · Troubleshooting

Qwen Image Edit Plus 2509 LoRA Inference with Diffusers (AI Toolkit-trained)

API model id: qwen_image_edit_plus
URL slug: qwen-image-edit-plus

This page documents the reference Diffusers inference pipeline for qwen_image_edit_plus (Qwen Image Edit Plus 2509). It is designed for running LoRAs trained with ostris/ai-toolkit while minimizing training preview vs inference mismatch. If you are trying to reproduce AI Toolkit sample previews, treat the code linked below as the source of truth (scheduler wiring, resolution snapping, LoRA application, and conditioning).

Run in the cloud (optional): If you want to reproduce the examples on this page in a pinned runtime without local CUDA/driver setup (and reduce preview‑vs‑inference drift), run it via RunComfy’s Cloud AI Toolkit (Train + Inference). 👉 You can open it here: Cloud AI Toolkit (Train + Inference)

Quick facts

Field	Value
Pipeline	`src/pipelines/qwen_image.py`
Base checkpoint	`Qwen/Qwen-Image-Edit-2509`
Defaults	`sample_steps=25`, `guidance_scale=4.0`, `seed=42`
Resolution snapping	Floors width/height to a multiple of 32
Control image	Required (`ctrl_img_1..3`)
LoRA scale behavior	Dynamic via adapters (`set_adapters`). Scale is set per request via `loras[].network_multiplier`.
Needs AI Toolkit	No

Reference implementation (source of truth)

Pipeline implementation: src/pipelines/qwen_image.py
Shared behaviors (snapping, seeding, LoRA base logic): src/pipelines/base.py
Request schema (parameter names): src/schemas/request.py
Model ids (enum): src/schemas/models.py
Pipeline registry (model → class): src/pipelines/__init__.py
Control-image loading (base64/URL): src/libs/image_utils.py
Executor (prompt processing + caching): src/tasks/executor.py

Minimal API request

{
  "model": "qwen_image_edit_plus",
  "trigger_word": "sks",
  "prompts": [
    {
      "prompt": "[trigger] a photo of a person",
      "width": 1024,
      "height": 1024,
      "seed": 42,
      "sample_steps": 25,
      "guidance_scale": 4.0,
      "neg": "",
      "ctrl_img_1": "<base64_or_url>",
      "ctrl_img_2": "<optional>",
      "ctrl_img_3": "<optional>"
    }
  ],
  "loras": [
    {
      "path": "my_lora_job/my_lora.safetensors",
      "network_multiplier": 1.0
    }
  ]
}

Control images

This model expects 1–3 reference images.

In this server implementation, provide them as:

ctrl_img_1 (required)
ctrl_img_2 (optional)
ctrl_img_3 (optional)

Notes:

The order of images matters.
For preview matching, keep the same preprocessing (size + resize kernel) as the reference pipeline.

Pipeline behavior that matters

Multi-image editing pipeline (diffusers.QwenImageEditPlusPipeline).
The server expects 1–3 reference images via ctrl_img_1..3 (not ctrl_img).
Prompt embeddings are encoded with preprocessed reference images using the pipeline’s CONDITION_IMAGE_SIZE.
Guidance uses true_cfg_scale.

Preview-matching notes (training preview vs inference mismatch)

Provide reference images as ctrl_img_1 (required), ctrl_img_2, ctrl_img_3. This implementation treats ctrl_img_1..3 as the authoritative list.
Changing the reference-image preprocessing (size/resize mode) is the fastest way to create preview mismatch.
Width/height are floored to a multiple of 32.

What to compare when debugging mismatch

Confirm the effective width/height after snapping (the server floors to the divisor).
Match sample_steps and the scheduler family (FlowMatch / UniPC / DDPM differences matter).
Match guidance_scale semantics (some pipelines map it to a different internal parameter).
Match loras[].network_multiplier and whether LoRA scale is dynamic vs fused.
If a control image is involved, match the resize method and the exact image(s) used.