Add Voice Bench: SFT package browser + model launch script

Self-contained browser tool for exploring the Vox Day / Nuttall SFT
packages and generating scenes against the joint CPT+SFT model, plus a
portable launch script (start-model.sh) and setup README for running
the model server on a fresh machine (DGX Spark or CUDA/L40 box).

README.md

@@ -0,0 +1,78 @@
+# Voice Bench
+
+A browser tool for exploring the Vox Day / Christopher Nuttall SFT packages
+and generating scenes from them against the joint CPT+SFT model, including a
+STYLE-swap control for testing voice conditioning across authors.
+
+## What's here
+
+- `voice-bench.html` — the tool itself. It's a single self-contained file
+  with no install step: open it in a browser (double-click it, or
+  File > Open). It needs no server of its own — it just needs a model server
+  to talk to (see below).
+- `start-model.sh` — starts the model server (`llama-server`, from
+  [llama.cpp](https://github.com/ggml-org/llama.cpp)) that `voice-bench.html`
+  sends generation requests to.
+
+The 208 training/holdout packages themselves are already embedded inside
+`voice-bench.html` — nothing else needs downloading for browsing.
+
+## One-time setup (per machine)
+
+### 1. Get the model file
+
+Download `Qwen3-30B-A3B-VoxDay-Nuttall-SFT-ep2-Q8_0.gguf` (~31GB) from the
+team Google Drive.
+
+Memory needed: ~31GB for the weights plus ~3GB of KV cache at the full
+32768-token context, so ~35GB total. Confirmed working on a DGX Spark;
+should comfortably fit a 48GB L40 as well. If a machine has less than that,
+lower `CTX_SIZE` (below) before anything else.
+
+### 2. Build llama.cpp with CUDA support
+
+Needs an NVIDIA GPU with the CUDA toolkit and `cmake` already installed.
+
+```
+git clone https://github.com/ggml-org/llama.cpp
+cd llama.cpp
+cmake -B build -DGGML_CUDA=ON
+cmake --build build --config Release -j
+```
+
+This produces `llama.cpp/build/bin/llama-server`, which is what
+`start-model.sh` runs.
+
+### 3. Start the model server
+
+```
+MODEL_PATH=/path/to/Qwen3-30B-A3B-VoxDay-Nuttall-SFT-ep2-Q8_0.gguf \
+LLAMA_SERVER=/path/to/llama.cpp/build/bin/llama-server \
+./start-model.sh
+```
+
+Leave this running in a terminal — it's the process `voice-bench.html`
+talks to. If you'd rather not set environment variables every time, edit the
+defaults at the top of `start-model.sh` instead.
+
+### 4. Open voice-bench.html
+
+Open the file in a browser. The "Server" field in the top-right corner
+defaults to `http://127.0.0.1:8200`, which is correct if the model server is
+running on the same machine as the browser. If it's running on a different
+machine on your network, change it to that machine's address instead, e.g.
+`http://192.168.1.50:8200` — and make sure that machine's firewall allows
+inbound connections on the port (`sudo ufw allow from <your-subnet> to any
+port 8200 proto tcp` on Ubuntu).
+
+## Troubleshooting
+
+- **"SERVER UNREACHABLE"** in the page header — the model server isn't
+  running yet, is still loading the model (can take a minute for a 31GB
+  file), or the Server field points at the wrong address/port.
+- **Out of memory when starting the server** — re-run with a smaller
+  context, e.g. `CTX_SIZE=8192 ./start-model.sh`.
+- **Opening the HTML file from a typed path does something odd in
+  Firefox** — use File > Open (Ctrl+O) instead of typing the path into the
+  address bar, or paste the full `file:///home/you/path/voice-bench.html`
+  URL.

start-model.sh

@@ -0,0 +1,47 @@
+#!/bin/bash
+# Starts the llama-server that voice-bench.html talks to.
+#
+# Requires:
+#   - llama.cpp built with CUDA support (see README.md)
+#   - the Qwen3-30B-A3B-VoxDay-Nuttall-SFT-ep2-Q8_0.gguf file (from the team
+#     Google Drive)
+#
+# Configure by setting environment variables before running, e.g.:
+#   MODEL_PATH=/data/models/Qwen3-30B-A3B-VoxDay-Nuttall-SFT-ep2-Q8_0.gguf \
+#   LLAMA_SERVER=/opt/llama.cpp/build/bin/llama-server \
+#   ./start-model.sh
+# or just edit the defaults below.
+set -e
+
+MODEL_PATH="${MODEL_PATH:-$HOME/models/Qwen3-30B-A3B-VoxDay-Nuttall-SFT-ep2-Q8_0.gguf}"
+LLAMA_SERVER="${LLAMA_SERVER:-$HOME/llama.cpp/build/bin/llama-server}"
+HOST="${HOST:-0.0.0.0}"
+PORT="${PORT:-8200}"
+CTX_SIZE="${CTX_SIZE:-32768}"
+
+if [ ! -f "$MODEL_PATH" ]; then
+  echo "!!! Model not found at: $MODEL_PATH"
+  echo "    Download it from the team Google Drive, then either place it at"
+  echo "    that path or re-run with MODEL_PATH=/path/to/the/file.gguf"
+  exit 1
+fi
+if [ ! -x "$LLAMA_SERVER" ]; then
+  echo "!!! llama-server not found/executable at: $LLAMA_SERVER"
+  echo "    Build llama.cpp first -- see README.md -- or re-run with"
+  echo "    LLAMA_SERVER=/path/to/llama-server"
+  exit 1
+fi
+
+echo "Model:  $MODEL_PATH"
+echo "Server: $LLAMA_SERVER"
+echo "Listening on ${HOST}:${PORT}, context ${CTX_SIZE}"
+echo "(if this exits with an out-of-memory error, re-run with a smaller CTX_SIZE, e.g. CTX_SIZE=8192)"
+echo
+
+exec "$LLAMA_SERVER" \
+  --model "$MODEL_PATH" \
+  --host "$HOST" \
+  --port "$PORT" \
+  -ngl 999 \
+  --ctx-size "$CTX_SIZE" \
+  --threads 8