Qwen-Image-Edit-Rapid-AIO¶

Development line: project:qwen-image-edit-rapid-aio · thread qwen-image-edit-rapid-aio
Last event: 2025-12-29 · 2 dated since 2025-10-07 · Researched: 2026-09-04 · confidence: high

What it is¶

Qwen-Image-Edit-Rapid-AIO merges speed accelerators, VAE, and CLIP for Qwen-Image-Edit into one checkpoint for ComfyUI users.

• Text-to-image without an input image
• Text-guided image editing with multiple input images
• SFW and NSFW variants

The workflow runs in 4 steps at CFG 1; v18 weighs 28.4 GB per variant.

It works for fast local editing and text-to-image, but this is not an official Qwen release and the author will not publish new versions.

Development line¶

• 2025-10-07 — Public repository reference for Qwen-Image-Edit-Rapid-AIO. The original v1 combined Qwen-Image-Edit-2509 and 4-step Lightning v2.0.
• 2025-12-29 — v18 resources and workflow references for Qwen-Image-Edit-Rapid-AIO. On 2025-12-29, the repository recorded links to its v18 directory, a Qwen-Rapid-AIO JSON workflow, and a related GGUF distribution. These dated links confirm a later distribution and workflow reference, but they do not prove the exact changes, compatibility, or performance of v18.

What changed¶

• 2025-10-07: The early Rapid-AIO checkpoint appeared; original v1 combined Qwen-Image-Edit-2509 and 4-step Lightning v2.0.
• 2025-12-29: v18 split into separate SFW and NSFW variants; the GGUF mirror added SFW quantizations Q4_K and Q5_K.
• 2026-01-08: v19 released.
• 2026-01-24: v23 released. The author later stated that v19 offers better editing consistency and v23 follows complex prompts better, then stopped further updates.

How to use this¶

As of 2025-12-29, start evaluation from the v18 repository resources and the JSON workflow, and check the sources before relying on GGUF availability, compatibility, or quality claims.

1. Download the chosen Rapid-AIO checkpoint and put it in the ComfyUI checkpoints folder. — https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO
2. Import the bundled JSON workflow; it uses CheckpointLoaderSimple, TextEncodeQwenImageEditPlus, KSampler, and VAE Decode. — https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO/blob/main/Qwen-Rapid-AIO.json
3. Connect one or two source images to optional image inputs, or leave them empty for text-to-image; set the prompt in TextEncodeQwenImageEditPlus. — https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO/blob/main/Qwen-Rapid-AIO.json
4. Run with 4 steps, CFG 1, sampler sa_solver, and scheduler beta from the workflow. — https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO/blob/main/Qwen-Rapid-AIO.json
5. For a smaller download, use the matching GGUF variant and a compatible GGUF loader instead of the standard checkpoint loader. — https://huggingface.co/Novice25/Qwen-Image-Edit-Rapid-AIO-GGUF/tree/main

Best practices¶

• Start with the bundled workflow and its 4-step/CFG 1 settings; do not copy standard Qwen-Image-Edit parameters blindly. — https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO/blob/main/Qwen-Rapid-AIO.json
• For edits, describe the specific change and pass the source into the image input; for text-to-image, leave the image input empty. — https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO/blob/main/Qwen-Rapid-AIO.json
• Downscale large inputs first if issues appear: the official Qwen-Image-Edit workflow scales input down to roughly one million pixels to prevent quality loss on oversized images. — https://docs.comfy.org/tutorials/image/qwen/qwen-image-edit
• Choose v19 over v23 for consistent edits; v23 is useful when following complex instructions matters more. — https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO

Superseded by this¶

• 2025-12-29: v18 SFW GGUF Q4_K/Q5_K expanded the earlier v18 NSFW quantizations.
• 2026-01-24: v23 was the final published release, but it does not replace v19 when edit consistency matters more.

Still unknown¶

• Rapid-AIO is a community checkpoint, not an official Qwen-Image-Edit release; base model capabilities cannot be assumed for every Rapid-AIO merge.
• The repository log for the 2025-10-07 event does not preserve the original commit date of the first v1 commit; the model card confirms the v1 description, but its historical date remains the event date.
• The Hugging Face URL for GGUF redirects from Arunk25 to Novice25; this looks like an account rename or transfer, but the reason is unconfirmed.

Sources¶

Agent brief¶

• Subject: project:qwen-image-edit-rapid-aio, thread qwen-image-edit-rapid-aio, 2 dated events 2025-10-07 → 2025-12-29.
• Practical note: As of 2025-12-29, evaluate v18 repository resources and the JSON workflow first, and check sources before trusting GGUF availability, compatibility, or quality claims.
• Confidence: high. Dated supersedes above determine what is obsolete.