Update README.md

README.md

@@ -11,6 +11,7 @@ license: apache-2.0
 short_description: An AI-driven visual novel — a small LLM writes every scene
 models:
   - google/gemma-4-12b-it
+  - build-small-hackathon/Gemma-26B-A4B-VisualNovel-GGUF
   - Abiray/Anima-base-v1.0-GGUF
   - circlestone-labs/Anima
   - Bingsu/adetailer
@@ -24,10 +25,18 @@ An AI-driven visual novel where every scene is written live by a **small
 LLM**. The model narrates, voices a three-character cast, swaps
 expressions and backgrounds, and offers branching choices — through a
 plain-text tool-call protocol (`[TOOL: set_expression …]`) that any small
-instruct model can speak. No fine-tune, no function-calling API.
+instruct model can speak — no function-calling API, and no fine-tune needed
+to play (this hosted demo runs a stock instruct model). At home we go one
+step further with a purpose-built fine-tune for sharper prose — see below.
 
 **Track:** Thousand Token Wood · **Hackathon:** Build Small
 
+## ▶ Demo
+
+[![Watch the demo](https://img.youtube.com/vi/0adUvKaqERM/maxresdefault.jpg)](https://www.youtube.com/watch?v=0adUvKaqERM)
+
+Watch it on YouTube: https://www.youtube.com/watch?v=0adUvKaqERM
+
 ## How to play
 
 1. Press **Begin** on the title screen.
@@ -39,7 +48,7 @@ instruct model can speak. No fine-tune, no function-calling API.
 
 | Piece | On this Space (ZeroGPU) | At home |
 |---|---|---|
-| Scene writer | `google/gemma-4-12b-it` (bf16) via transformers, `@spaces.GPU` | Gemma 4 26B-A4B MoE Q4_K_M GGUF on stock llama.cpp with MTP speculative decoding, on an 8 GB GPU |
+| Scene writer | `google/gemma-4-12b-it` (bf16) via transformers, `@spaces.GPU` | Our **[Gemma 4 26B-A4B VN fine-tune](https://huggingface.co/build-small-hackathon/Gemma-26B-A4B-VisualNovel-GGUF)** — MoE, Q4_K_M GGUF on stock llama.cpp with MTP speculative decoding, on an 8 GB GPU |
 | Character sprites | Live ComfyUI pipeline, **embedded in-process** under `@spaces.GPU`: face-gated base gen → expression warping → background removal | Same pipeline, ComfyUI as a subprocess over HTTP |
 | Backgrounds | Preset library + live txt2img for novel keys (same embedded pipeline) | Preset + live txt2img for novel scene keys |
 
@@ -55,13 +64,4 @@ are downloaded at startup. On an 8 GB card the LLM and the diffusion stack
 can't co-reside, so the home setup stops/relaunches servers around image
 bakes; on ZeroGPU none of that is needed — the GPU is granted per call.
 There's also a Docker variant (see `Dockerfile` on the repo) that runs
-the full 26B + live ComfyUI casting on a dedicated T4.
-
-## Run it locally
-
-```bash
-pip install -r requirements.txt gradio==6.16.0
-bash tools/start_gemma_server_stock.sh &   # stock llama.cpp + MTP
-python app.py
-ARS_FABULA_BACKEND=mock python app.py      # no model needed
-```
+the full 26B + live ComfyUI casting on a dedicated T4.