Initial commit: ComfyUI-VideoToText

2026-05-22 09:33:52 +08:00
commit 7e0e1bf4fb
21 changed files with 1964 additions and 0 deletions
@@ -0,0 +1,103 @@
+#!/bin/bash
+#
+# install_server.sh — idempotent server-side install for ComfyUI-VideoToText.
+#
+# Usage:
+#     COMFYUI_DIR=/mnt/modelall/ComfyUI ./scripts/install_server.sh
+#
+# What it does:
+#     1. Verifies ffmpeg is on PATH
+#     2. Installs torchaudio into ComfyUI's venv (uv-managed)
+#     3. Smoke-tests the imports
+#
+# What it does NOT do:
+#     - Install MiMoASR (use that pack's own install_server.sh)
+#     - Install rgthree-comfy (clone it manually if needed)
+#     - Download model weights
+#     - Restart ComfyUI (you do this after; see README Step 7)
+#
+set -euo pipefail
+
+# ----- helpers -----
+log()  { printf '\033[1;34m[+]\033[0m %s\n' "$*"; }
+warn() { printf '\033[1;33m[!]\033[0m %s\n' "$*"; }
+err()  { printf '\033[1;31m[x]\033[0m %s\n' "$*" >&2; }
+
+# ----- 1. ffmpeg -----
+log "Checking ffmpeg ..."
+if ! command -v ffmpeg >/dev/null 2>&1; then
+    err "ffmpeg not on PATH. Install with:  sudo apt install ffmpeg"
+    exit 1
+fi
+ffmpeg -version | head -1
+
+# ----- 2. ComfyUI venv -----
+: "${COMFYUI_DIR:?Please set COMFYUI_DIR (e.g. /mnt/modelall/ComfyUI)}"
+if [[ ! -d "$COMFYUI_DIR/.venv" ]]; then
+    err "ComfyUI venv not found: $COMFYUI_DIR/.venv"
+    exit 1
+fi
+
+log "Activating ComfyUI venv ..."
+# shellcheck disable=SC1091
+source "$COMFYUI_DIR/.venv/bin/activate"
+
+# Pick uv if available; fall back to pip
+if command -v uv >/dev/null 2>&1; then
+    PIP="uv pip"
+else
+    PIP="pip"
+fi
+
+# ----- 3. torchaudio -----
+log "Installing torchaudio ..."
+$PIP install torchaudio
+
+log "Smoke-test imports ..."
+python - <<'PY'
+import sys
+ok = True
+try:
+    import torchaudio
+    print(f"  torchaudio {torchaudio.__version__} OK")
+except Exception as e:
+    print(f"  torchaudio FAILED: {e}"); ok = False
+
+try:
+    import folder_paths  # ComfyUI core module
+    print(f"  folder_paths OK ({folder_paths.__file__})")
+except Exception as e:
+    print(f"  folder_paths (ComfyUI core): cannot import outside ComfyUI runtime — that's normal, skipping")
+
+sys.exit(0 if ok else 1)
+PY
+
+# ----- 4. final advice -----
+cat <<'NEXT'
+
+[+] torchaudio installed.
+
+Next steps (manual):
+  1. If not yet installed: clone ComfyUI-MiMoASR and rgthree-comfy
+     $ cd "$COMFYUI_DIR/custom_nodes"
+     $ git clone https://github.com/aadebuger/ComfyUI-MiMoASR.git
+     $ git clone https://github.com/rgthree/rgthree-comfy.git
+
+  2. Download MiMo weights (~24 GB):
+     $ export HF_HOME=/mnt/modelall/hf_cache
+     $ hf download XiaomiMiMo/MiMo-V2.5-ASR --local-dir /mnt/modelall/models/MiMo-V2.5-ASR
+     $ hf download XiaomiMiMo/MiMo-Audio-Tokenizer --local-dir /mnt/modelall/models/MiMo-Audio-Tokenizer
+
+  3. Clone MiMo upstream repo:
+     $ git clone https://github.com/XiaomiMiMo/MiMo-V2.5-ASR.git /opt/repos/MiMo-V2.5-ASR
+
+  4. Restart ComfyUI with required env vars:
+     $ pkill -9 -f main.py; sleep 3
+     $ export HF_HOME=/mnt/modelall/hf_cache
+     $ export MIMO_ASR_REPO=/opt/repos/MiMo-V2.5-ASR
+     $ cd "$COMFYUI_DIR" && source .venv/bin/activate
+     $ nohup python main.py --listen 0.0.0.0 --port 8188 > ~/comfy_start.log 2>&1 &
+
+  5. Verify 9 nodes registered (see README Step 8).
+
+NEXT