qwen/qwen-image-lora-trainer 🔢🖼️📝✓❓ → 🖼️

▶️ 454 runs 📅 Nov 2025 ⚙️ Cog 0.16.9 🔗 GitHub 📄 Paper

About

Fine-tunable Qwen Image model with exceptional composition abilities - train custom LoRAs for any style or subject

Input Parameters

seed Type: integer: Random seed. Leave blank for random.
image Type: string: Optional guide image for img2img.
prompt (required) Type: string: Prompt for generated image
go_fast Type: booleanDefault: true: Run faster predictions with aggressive caching.
guidance Type: numberDefault: 3Range: 0 - 10: Guidance for generated image (0-10).
strength Type: numberDefault: 0.9Range: 0 - 1: Strength for img2img pipeline
image_size Default: optimize_for_quality: Image size preset (quality = larger, speed = faster).
lora_scale Type: numberDefault: 1: Determines how strongly the loaded LoRA should be applied.
aspect_ratio Default: 16:9: Aspect ratio for the generated image
lora_weights Type: string: Load LoRA weights. Supports local .safetensors paths or ZIPs produced by cog train.
output_format Default: webp: Format of the output images
enhance_prompt Type: booleanDefault: false: Append a high-detail suffix to the prompt.
output_quality Type: integerDefault: 80Range: 0 - 100: Quality when saving lossy images (0-100).
negative_prompt Type: stringDefault:: Negative prompt for generated image
replicate_weights Type: string: LoRA ZIP generated by cog train (alternate to lora_weights).
num_inference_steps Type: integerDefault: 30Range: 1 - 50: Number of denoising steps (1–50).
disable_safety_checker Type: booleanDefault: false: Disable safety checker (not used).

Output Schema

Output

Type: array • Items Type: string • Items Format: uri

Version Details

Version ID: f28eb39544f2c0dff4fbd9d50588fd75789f7ef26f6118456a96c2eedddddf90
Version Created: November 15, 2025