LongCat-Image edit

[Talmaj]

Repackaged model: https://huggingface.co/TalmajM/LongCat-Image-Edit_ComfyUI_repackaged/tree/main
Workflow: longcat_image_edit final.json

The outputs differ slightly from the original, because we pre-process the input image using existing comfy nodes.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 4dc4ffb4-6113-451d-b95f-d95adaeb37f7

📥 Commits

Reviewing files that changed from the base of the PR and between 7a16e8a and 7174b95.

📒 Files selected for processing (5)

• comfy/ldm/flux/model.py
• comfy/model_base.py
• comfy/text_encoders/llama.py
• comfy/text_encoders/longcat_image.py
• comfy/text_encoders/qwen_vl.py

---

📝 Walkthrough

Walkthrough

The PR modifies five files across text encoding and model architecture components. Changes include: propagating transformer_options through the image patch processing pipeline in the Flux model; adjusting rope-shift configuration to derive from cross-attention context length instead of fixed values; modifying position ID calculations in Qwen text encoding to account for attention masks; extending the LongCat image tokenizer with image padding constants and updating its tokenization method to handle image parameters; and adding a reordering step in Qwen VL's forward pass to restore windowed sequence ordering.

🚥 Pre-merge checks | ✅ 1 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Title check	❓ Inconclusive	The title 'LongCat-Image edit' is vague and does not clearly describe the specific changes made across multiple files (flux model, rope options, position handling, tokenizer updates).	Consider a more descriptive title that captures the main technical changes, such as 'Add LongCat-Image edit support with position handling and tokenizer updates' or similar.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Description check	✅ Passed	The description provides context about the repackaged model and workflow file, which relates to the changeset's purpose of enabling LongCat-Image edit functionality.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

📝 Coding Plan

• Generate coding plan for human review comments

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Tip

CodeRabbit can use Trivy to scan for security misconfigurations and secrets in Infrastructure as Code files.

Add a .trivyignore file to your project to customize which findings Trivy reports.

[Talmaj]

The output still appears nice without having this change. However, the numerical precision (cosine distance) between the original implementation and this one jumps by a significant amount when adding this part.

[yu234567]

Does it support turbo?LongCat-Image-Edit-Turbo

[Talmaj]

Does it support turbo?LongCat-Image-Edit-Turbo

I haven't tested it, but it should work since it's using the same workflow/pipeline. I've repackaged the turbo model here:
https://huggingface.co/TalmajM/LongCat-Image-Edit-Turbo_ComfyUI_repackaged

Modify the provided workflow to use this model instead.

[zwukong]

Thanks , how it the fp8 quality, fp8 smaller and faster normally

[slmonker]

Is it still a single image edit model？

[Amazon90]

Has anyone tried Longcat image edit turbo? Compared to the non‑turbo version, which one is actually worth keeping?

[slmonker]

Has anyone tried Longcat image edit turbo? Compared to the non‑turbo version, which one is actually worth keeping?

Minor difference, about four times faster.