Picture published by the creator on Hugging Face.
humanizer: Rewrite AI-written drafts so they read like a person (12B)
Added · Updated
Made by Stephen Yu
View profile on Hugging Face (opens in a new tab)Paste an AI-written email, essay or report and get a rewrite that reads more naturally, offline; for people writing in English or Chinese.
Model details
Original model jialinyyzz/humanizer
Input Text Output Text
Ways to use it Run locally
5 reviewed local configurations. Add My computer details to compare them.
Q4_K_M GGUF on Apple silicon · llama.cpp · The pinned card documents llama.cpp on macOS, Windows, and Linux, plus a Mac Apple silicon app. This configuration uses the documented Mac app path.
How this configuration was checked
16 GB memory · reviewed estimate Source lists 7.6 GB (file-size). Estimate uses 7.63 GB exact GGUF file size, 0.269 GB 4K KV cache (8 full-attention layers from 48 hybrid layers × 8 KV heads × 256 head width × K/V × 2-byte values), 4 GB llama.cpp runtime buffers, and 4.101 GB remaining in the 16 GB unified-memory tier for macOS and apps.
quantized · Q4_K_M · 4,096 token context · source context 8,192 tokens · batch size 1 · concurrency 1 · Single-user 4K comparison; the card recommends an 8K context for its documented rewrite workflow.
Source checked 2026-10-21 · revision 2e57b6528a4070c33f3e09ca24e3325a84272f68. pinned model card lists the exact Q4_K_M GGUF size, llama.cpp platforms, and 8K context; threshold is estimated by the fixed file-size and config formula
Role in this family: Fine-tuned version.
google/gemma-4-12B (opens in a new tab)What it does
humanizer is a 12B model, fine-tuned from Google's Gemma 4 12B, that rewrites AI-written drafts so they read like a person wrote them. The source says it is trained to keep every number, unit, date, name and quote and to add nothing. It runs locally, and the source offers several file sizes, including smaller compressed versions.
How to use it
Input: a draft such as an email, essay, report or forum post, in English or Chinese. Action: download the desktop app, open it and paste the draft; the source says it works offline. You can also run the model files without the app using llama.cpp, following the usage guide. Output: a rewritten version of the draft.
Common use cases
- Paste an AI-written email and get a rewrite meant to read like a person wrote it; the model is trained to keep numbers and names, so check them.
- Rewrite a report or essay draft in English or Chinese, then compare the dates and quotes against your original.
- Rewrite a forum post or answer before posting it.
Can I use this?
Requirements Download the Mac (Apple silicon) .dmg or the Windows installer from the project's releases, or follow the guide for running without the app. The source says the app can import .docx and PDF files. The license is Apache-2.0. RAM, GPU and exact disk needs are not documented.
Caveat The creator's own strict judge found no factual problem in 376 of 420 English rewrites, so some rewrites still contain errors, mostly a single word or phrase. Check every result, especially numbers, dates and names, before use.
Worth knowing The source says to read the result before you send it, especially numbers, dates and names. Its own judge flagged 44 of the 420 English rewrites.
Why it's here
Not trending this week
Last seen on the Hugging Face trending list on 22 October 2026. It stays on BuildTube so you can still find it.
- 2,253 people have liked it on Hugging Face.
- It was downloaded 369,624 times in the last 30 days.
Numbers checked on 8 November 2026; not refreshed since.
Behind it
View model pageBuildTube has not run or verified this project. Everything above is written from what the creator published.
Made this?
Quick questions
- What is humanizer?
- humanizer is a 12B model, fine-tuned from Google's Gemma 4 12B, that rewrites AI-written drafts so they read like a person wrote them.
- What can I use humanizer for?
- Paste an AI-written email and get a rewrite meant to read like a person wrote it; the model is trained to keep numbers and names, so check them; Rewrite a report or essay draft in English or Chinese, then compare the dates and quotes against your original; Rewrite a forum post or answer before posting it
- Can I run humanizer on my computer?
- It runs on your own hardware with a model runtime. Download the Mac (Apple silicon) .dmg or the Windows installer from the project's releases, or follow the guide for running without the app. RAM, GPU and exact disk needs are not documented.
- What do I need to use humanizer?
- The app is offered for Mac with Apple silicon and for Windows; the model also runs without the app through llama.cpp.
- What license does humanizer use?
- The license is Apache-2.0.
More from this week
- P-H3-Image-Comfy Make and edit 8-megapixel images in ComfyUI
Turn short text requests into 8-megapixel images and edit pictures on your own graphics card.
- Idol Turn one picture of an object into a 3D Gaussian splat
Idol turns one photo, render or generated image of an object into a 3D Gaussian splat file.
- agent-desktop Let AI agents read and click desktop apps
For developers building computer-use agents: agent-desktop lets an AI agent read and operate desktop apps on Mac, Windows or Linux from a command line.
Explore by topic
Post this
Download share imageFor makers
Made this? Show it’s on BuildTube — add the badge to your README or embed the card on your site.
README badge (Markdown)
[](https://buildtube.app/p/hf-jialinyyzz-humanizer/)Embed card (HTML)
<iframe src="https://buildtube.app/embed/hf-jialinyyzz-humanizer/" width="560" height="240" style="border:0;max-width:100%" loading="lazy" title="humanizer on BuildTube"></iframe>