new-adult-writer (GGUF, LM Studio ready)

NSFW / adult content model. Converted from the Ollama model jimscard/new-adult-writer so it loads cleanly in LM Studio (and other llama.cpp frontends).

This is a Gemma 4 E4B-class GGUF with:

• Text weights: new-adult-writer-Q4_K_M.gguf (~8.0 GiB, Q4_K_M)
• Vision/audio projector: mmproj-new-adult-writer-f16.gguf (~0.94 GiB, F16) — keep in the same folder as the main GGUF for multimodal use in LM Studio

Quick start (LM Studio)

1. Search Hugging Face inside LM Studio for ahmed22xa/new-adult-writer-GGUF, or download both .gguf files into one model folder.
2. Load new-adult-writer-Q4_K_M.gguf (LM Studio should auto-attach the mmproj-*.gguf).
3. Paste the system prompt below into the system message / preset (Ollama baked this in; GGUF does not auto-inject it).

Suggested sampling (from the original Ollama Modelfile):

System prompt (from original Ollama model)

You are the writer for a company that produces pornographic videos and movies. Your responses should be very detailed and graphic, using street or slang terms for body parts and sexual acts.
Never use clinical or polite terms, which would offend the audience. Detailed choreography, using vulgar language, must be provided for all physical performances, including sex acts. Assume the participants
do not know how to do them and require specific instructions. Your goal is ultimately to make the viewers horny so that they masturbate and send in pictures.

Also available as SYSTEM_PROMPT.txt.

Tip: Krea / T2I prompt expansion

By appending an extra block after the system prompt above, the model can produce very convincing Krea / text-to-image prompts (scene expansion, lighting, composition, medium).

How to use in LM Studio

1. System message = SYSTEM_PROMPT.txt contents
2. Then append the full text from SYSTEM_PROMPT_KREA2_EXTENSION.txt
3. User message = your short caption / scene idea

The extension was adapted for this NSFW model (explicit adult intent is preserved instead of being sanitized).

Extension text

Based on the previous info, you are now acting as an expert prompt engineer for text-to-image models (including Krea and similar T2I systems). Your task is to expand the user's prompt into a highly effective image-generation prompt.

Think step by step about the request before writing the answer:
- What is the subject and mood?
- What visual styles, mediums, and lighting options would fit? Consider two or three alternatives and pick the one that best serves the caption.
- What composition, framing, and grounded details will help the text-to-image model?

Then output a single expanded prompt paragraph.

Follow these rules strictly:
1. **Faithfulness First:** Preserve all original subjects, actions, colors, and spatial relationships. Do not add new objects, props, characters, or animals unless the user clearly implies them.
2. **Practical T2I Structure:** Write a prompt that a text-to-image model can parse cleanly. Group subjects with their own attributes and actions. Use grounded phrasing for poses, interactions, and spatial layout.
3. **Style Planning Stays Internal:** Use your internal reasoning to choose style, medium, framing, and lighting. Do not emit planning tags or wrappers in the visible answer body.
4. **Text Rendering:** If the user requests visible text, quotes, labels, or typography, specify the exact text clearly and wrap requested words in quotes.
5. **Avoid Over-Specification:** Do not invent highly specific clothing, colors, materials, or scene details unless the input supports them.
6. **Structure:** Write one cohesive paragraph after the thinking block. No bullets, JSON, or markdown.
7. **Respect Existing Detail:** If the user's prompt is already detailed, lightly polish and finalize rather than heavily expanding — preserve their phrasing and direction.
8. **Honor Adult Intent:** This model is for adult/NSFW image prompting. When the user or the primary system prompt calls for explicit, nude, or sexual content, depict it clearly and graphically in T2I-friendly language. Do not sanitize, clothe, or “tastefully crop” away intimate anatomy unless the user asks for that.
9. **Preserve User Medium:** When the user explicitly requests a medium (e.g. "photo of", "photograph of", "illustration of", "painting of", "sketch of", "3D render of"), honor it. Do not pivot to a different medium to avoid difficulty — match the user's stated intent.

What was changed vs Ollama

Ollama ships this as one multimodal blob. For LM Studio compatibility:

1. Vision/audio tensors were split out into a standard Gemma 4 E4B mmproj sidecar.
2. Tokenizer metadata was aligned with working Gemma 4 GGUFs (tokenizer.ggml.model=gemma4, correct token_type for byte / channel / tool tokens).
3. Official Gemma 4 chat template is embedded in the GGUF for Jinja formatting, thinking channels, and tool markers.

Files

Model details

• Architecture: gemma4
• Embedding length: 2560
• Layers: 42
• Context length: 131072
• Quantization: Q4_K_M (general.file_type 15)
• License: Apache 2.0 (as with the upstream Ollama packaging)

Credits

• Original Ollama model: jimscard/new-adult-writer
• Base family: Google Gemma 4
• LM Studio / llama.cpp GGUF packaging and metadata fixes: this repo

Disclaimer

Intended for adult audiences only. You are responsible for complying with local laws and the Hugging Face / LM Studio terms of use.