refactor(cli): defer soundfile and voxcpm.core imports to inference commands

Move `import soundfile as sf` and `from voxcpm.core import VoxCPM` from
module-level into the functions that require model inference (load_model,
_run_single, cmd_batch), so `voxcpm validate` can run without loading
the model/inference stack.

src/voxcpm/cli.py

@@ -11,11 +11,6 @@ import os
 import sys
 from pathlib import Path
 
-import soundfile as sf
-
-from voxcpm.core import VoxCPM
-
-
 DEFAULT_HF_MODEL_ID = "openbmb/VoxCPM2"
 
 # -----------------------------
@@ -173,7 +168,9 @@ def validate_batch_args(args, parser):
 # -----------------------------
 
 
-def load_model(args) -> VoxCPM:
+def load_model(args):
+    from voxcpm.core import VoxCPM
+
     print("Loading VoxCPM model...", file=sys.stderr)
 
     zipenhancer_path = getattr(args, "zipenhancer_path", None) or os.environ.get(
@@ -266,6 +263,8 @@ def _run_single(args, parser, *, text: str, output: str, prompt_text: str | None
         and (args.prompt_audio is not None or args.reference_audio is not None),
     )
 
+    import soundfile as sf
+
     sf.write(str(output_path), audio_array, model.tts_model.sample_rate)
 
     duration = len(audio_array) / model.tts_model.sample_rate
@@ -307,6 +306,8 @@ def cmd_validate(args, parser):
 
 
 def cmd_batch(args, parser):
+    import soundfile as sf
+
     input_file = require_file_exists(args.input, parser, "input file")
     output_dir = Path(args.output_dir)
     output_dir.mkdir(parents=True, exist_ok=True)