Banana_Styles Styler (Advanced)
Yes, the demo menu has an SDXL dual-encoder version too
- text_positive_g
- text_positive_l
- text_positive
- text_negative_g
- text_negative_l
- text_negative
Yes, there's an Advanced version of the banana node. That tells you something reassuring about MileHighStyler: it doesn't hand-write nodes, it generates them, so every menu automatically gets both a basic and an "Advanced" flavor - even the ten-banana demo menu. Which is exactly why this one is worth a minute. It's the least intimidating possible place to learn what "Advanced" means in this pack, because the styles themselves are a joke and there's nothing at stake.
What "Advanced" actually means
SDXL runs two text encoders - a big OpenCLIP-G and a smaller CLIP-L - side by side. The Advanced nodes expose that split. Instead of one positive box you get text_positive_g and text_positive_l, one per encoder, and you can feed them different text if you want (broad idea to G, fine detail to L) or the same text to both. Then negative_prompt_to lets you send your negative to Both, G only, or L only.
That's the entire difference. On the banana menu it's obviously pointless, but the wiring is identical to the serious Advanced stylers - so if you can drive this, you can drive CameraStylerAdvanced or the Azazeal one. Note that the split only exists on SDXL; on SD 1.5 there's one encoder, and on 2026 LLM-encoded models (Flux 2, Z-Image, Qwen) the CLIP-G/CLIP-L concept is gone entirely.
How it works
It's a fork of SDXL Prompt Styler. Each banana entry is a JSON template with a {prompt} slot; your text drops in, the surrounding words get wrapped at a strong weight (~2.0), and the node outputs rewritten prompt strings. Pure text - nothing loads a model.
Inputs and outputs that matter
- banana_styles - the ten-entry dropdown.
- text_positive_g / text_positive_l - your subject, once per SDXL encoder.
- negative_prompt_to -
Both(default),G only, orL only.
log_prompt prints the built prompt. The node returns six strings - text_positive_g, text_positive_l, combined text_positive, and the three negative equivalents. Route _g/_l to an SDXL dual CLIP Text Encode, or use the combined text_positive / text_negative and ignore the split.
Installing it
ComfyUI Manager: search ComfyUI_MileHighStyler, install, restart. Or:
cd ComfyUI/custom_nodes && git clone https://github.com/TripleHeadedMonkey/ComfyUI_MileHighStyler
No models, no dependencies. Nodes appear under ali1234/stylers.
Where people get burned
- Expecting the banana menu to be different because it's Advanced. It's the same ten bananas. "Advanced" is about the encoder wiring, not better styles.
- Splitting G and L on a demo. There's no reason to. If you're just learning the sockets, put the same text in both and watch the log.
- Carrying it to a non-SDXL model. The extra sockets do nothing off SDXL. Use the plain node, or a menu that isn't fruit.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| text_positive_g | STRING | — | |
| text_positive_l | STRING | — | |
| text_negative | STRING | — | |
| banana_styles | COMBO | 10 options: Standard Yellow Banana, Ripe and Sweet Banana, Organic Banana, Plantain, Red Banana, Cavendish Banana, +4 | |
| negative_prompt_to | COMBO | Both | 3 options: Both, G only, L only |
| log_prompt | BOOLEAN | true | — |
Outputs (6)
| Name | Type | Description |
|---|---|---|
| text_positive_g | STRING | — |
| text_positive_l | STRING | — |
| text_positive | STRING | — |
| text_negative_g | STRING | — |
| text_negative_l | STRING | — |
| text_negative | STRING | — |