WO_JoyCaption_ExtraOptions

The main JoyCaption node keeps its interface clean by pushing all the tuning over here. This node is just a control panel: twenty-seven boolean toggles plus a character_name field, and its only job is to produce the extra_options output that you plug into the main node's extra_options input. Turn nothing on and the workflow behaves exactly as if the node weren't there - everything defaults to off.

Understand what it's doing and it stops feeling like a wall of checkboxes. Each toggle appends an instruction sentence to the query that gets sent to the model. These aren't model knobs; they're prompt engineering with a UI. Pick a few relevant ones rather than flipping everything on, because each one is competing for the model's attention.

The ones that matter most

• refer_character_name plus character_name - this is the identity-anchoring pair. If there's a person in the image, the model must refer to them as {name}. For a character or person LoRA this is the difference between "the model describes your character as a generic man" and "the model names your character," which is exactly the describe-the-variable, leave-the-fixed rule from the LoRA training playbook.
• exclude_people_info - drop unchangeable traits (ethnicity, gender) but keep changeable ones like hairstyle. The textbook captioning correction for over-describing fixed features.
• avoid_meta_descriptive_phrases - suppresses useless "This image shows..." framing, which matters when you're captioning for a text-to-image model.
• include_exif - forces photo-detail captioning: likely camera, aperture, shutter speed, ISO. Gold for photoreal training sets.
• The NSFW cluster: use_vulgar_slang_and_profanity, include_nsfw_rating, do_not_use_polite_euphemisms, and exclude_sexual. This is the uncensored-captioner feature, and it's the thing most API captioners physically cannot do.

Then there's the long tail - include_lighting, include_camera_angle, include_camera_shot_type, include_camera_vantage_height, specify_depth_field, include_composition_style, include_aesthetic_quality, include_watermark_info / mention_watermark_explicitly, include_JPEG_artifacts, exclude_text, exclude_image_resolution, exclude_mood_feeling, identify_image_orientation, only_describe_most_important_elements, and a few more. Read the tooltip on each - the author wrote them as actual instructions, and they tell you exactly what gets added.

The one thing to watch

This node only feeds the main JoyCaption node. The JoyCaption (Custom) node has no extra_options input, so all these toggles do nothing there - if you're on Custom, you'd have to type the instructions into your query yourself. Wiring is otherwise trivial: extra_options out, straight into the main node's extra_options in. Install it the same way as the rest of the pack - ComfyUI Manager search wo_joycaption_comfyui, or git clone https://github.com/without-ordinary/wo_joycaption_comfyui into custom_nodes.