feat(minimax tts): expose all voice/audio parameters on the tts subcommand

minimax/agent-harness/cli_anything/minimax/README.md

@@ -56,10 +56,32 @@ cli-anything-minimax tts --text "Hello, world!" --output hello.mp3
 # Use turbo model
 cli-anything-minimax tts --text "Fast speech" --model speech-2.8-turbo --output fast.mp3
 
+# Fine-grained voice & audio control
+cli-anything-minimax tts --text "Slow and quiet" \
+    --speed 0.8 --vol 0.5 --pitch -2 \
+    --sample-rate 44100 --bitrate 256000 --format flac --channel 2 \
+    --output slow.flac
+
 # List available voices
 cli-anything-minimax voices
 ```
 
+#### TTS options
+
+| Option | Range / Choices | Default | Description |
+|--------|-----------------|---------|-------------|
+| `--text` / `-t` | (required) | — | Text to synthesize |
+| `--model` | model id | `speech-2.8-hd` | TTS model |
+| `--voice` | voice id | `English_Graceful_Lady` | Voice preset |
+| `--output` / `-o` | path | `output.mp3` | Output audio file |
+| `--speed` | 0.5 .. 2.0 | `1.0` | Speech speed multiplier |
+| `--vol` | 0.0 .. 10.0 | `1.0` | Volume |
+| `--pitch` | -12 .. 12 | `0` | Pitch shift in semitones |
+| `--sample-rate` | 8000 / 16000 / 22050 / 24000 / 32000 / 44100 | `32000` | Audio sample rate |
+| `--bitrate` | 32000 / 64000 / 128000 / 256000 | `128000` | Audio bitrate |
+| `--format` | mp3 / pcm / flac | `mp3` | Output container |
+| `--channel` | 1 / 2 | `1` | 1 = mono, 2 = stereo |
+
 ### Session & Config
 
 ```bash

minimax/agent-harness/cli_anything/minimax/minimax_cli.py

@@ -256,9 +256,59 @@ def on_chunk(chunk_content):
     show_default=True,
     help="Output audio file path",
 )
+@click.option(
+    "--speed",
+    type=click.FloatRange(0.5, 2.0),
+    default=1.0,
+    show_default=True,
+    help="Speech speed (0.5-2.0).",
+)
+@click.option(
+    "--vol",
+    type=click.FloatRange(0.0, 10.0),
+    default=1.0,
+    show_default=True,
+    help="Volume (0.0-10.0).",
+)
+@click.option(
+    "--pitch",
+    type=click.IntRange(-12, 12),
+    default=0,
+    show_default=True,
+    help="Pitch shift in semitones (-12..12).",
+)
+@click.option(
+    "--sample-rate",
+    type=click.Choice(["8000", "16000", "22050", "24000", "32000", "44100"]),
+    default="32000",
+    show_default=True,
+    help="Audio sample rate.",
+)
+@click.option(
+    "--bitrate",
+    type=click.Choice(["32000", "64000", "128000", "256000"]),
+    default="128000",
+    show_default=True,
+    help="Audio bitrate.",
+)
+@click.option(
+    "--format",
+    "audio_format",
+    type=click.Choice(["mp3", "pcm", "flac"]),
+    default="mp3",
+    show_default=True,
+    help="Output audio format.",
+)
+@click.option(
+    "--channel",
+    type=click.Choice(["1", "2"]),
+    default="1",
+    show_default=True,
+    help="1=mono, 2=stereo.",
+)
 @click.pass_context
 @handle_error
-def tts(ctx, text, model_opt, voice, output_path):
+def tts(ctx, text, model_opt, voice, output_path, speed, vol, pitch, sample_rate, bitrate, audio_format, channel):
     """Synthesize text to speech using MiniMax TTS."""
     parent_key = ctx.obj.get("api_key") if ctx.obj else None
     api_key = get_api_key(parent_key)
@@ -269,6 +319,13 @@ def tts(ctx, text, model_opt, voice, output_path):
         model=model_opt,
         voice=voice,
         output_path=output_path,
+        speed=speed,
+        vol=vol,
+        pitch=pitch,
+        sample_rate=int(sample_rate),
+        bitrate=int(bitrate),
+        audio_format=audio_format,
+        channel=int(channel),
     )
 
     output_data = {

minimax/agent-harness/cli_anything/minimax/tests/test_tts_extended.py

@@ -0,0 +1,78 @@
+"""Tests for the extended tts_synthesize parameters (speed / vol / pitch / audio settings).
+
+The new params were previously hardcoded inside tts_synthesize and only the
+backend payload used them. These tests guard the new public surface.
+"""
+
+import json
+from unittest.mock import patch, MagicMock
+
+from cli_anything.minimax.utils.minimax_backend import tts_synthesize
+
+
+def _empty_sse_post():
+    """Build a requests.post mock that returns an empty SSE stream."""
+    mock_post = MagicMock()
+    mock_resp = MagicMock()
+    mock_resp.status_code = 200
+    mock_resp.raise_for_status = lambda: None
+    mock_resp.iter_content.return_value = []
+    mock_post.return_value = mock_resp
+    return mock_post
+
+
+def _captured_payload(mock_post):
+    return mock_post.call_args.kwargs["json"]
+
+
+def test_tts_default_voice_audio_settings():
+    with patch("requests.post", _empty_sse_post()) as mock_post:
+        tts_synthesize(api_key="key", text="hi")
+        payload = _captured_payload(mock_post)
+
+    assert payload["voice_setting"]["speed"] == 1.0
+    assert payload["voice_setting"]["vol"] == 1.0
+    assert payload["voice_setting"]["pitch"] == 0
+    assert payload["audio_setting"]["sample_rate"] == 32000
+    assert payload["audio_setting"]["bitrate"] == 128000
+    assert payload["audio_setting"]["format"] == "mp3"
+    assert payload["audio_setting"]["channel"] == 1
+
+
+def test_tts_custom_voice_setting():
+    with patch("requests.post", _empty_sse_post()) as mock_post:
+        tts_synthesize(api_key="key", text="hi", speed=1.5, vol=3.0, pitch=4)
+        payload = _captured_payload(mock_post)
+
+    assert payload["voice_setting"]["speed"] == 1.5
+    assert payload["voice_setting"]["vol"] == 3.0
+    assert payload["voice_setting"]["pitch"] == 4
+
+
+def test_tts_custom_audio_setting():
+    with patch("requests.post", _empty_sse_post()) as mock_post:
+        tts_synthesize(
+            api_key="key",
+            text="hi",
+            sample_rate=44100,
+            bitrate=256000,
+            audio_format="flac",
+            channel=2,
+        )
+        payload = _captured_payload(mock_post)
+
+    assert payload["audio_setting"]["sample_rate"] == 44100
+    assert payload["audio_setting"]["bitrate"] == 256000
+    assert payload["audio_setting"]["format"] == "flac"
+    assert payload["audio_setting"]["channel"] == 2
+
+
+def test_tts_voice_id_propagates():
+    with patch("requests.post", _empty_sse_post()) as mock_post:
+        tts_synthesize(
+            api_key="key", text="hi", voice="English_Lucky_Robot", speed=0.8
+        )
+        payload = _captured_payload(mock_post)
+
+    assert payload["voice_setting"]["voice_id"] == "English_Lucky_Robot"
+    assert payload["voice_setting"]["speed"] == 0.8

minimax/agent-harness/cli_anything/minimax/utils/minimax_backend.py

@@ -183,6 +183,13 @@ def tts_synthesize(
     model: str = "speech-2.8-hd",
     voice: str = "English_Graceful_Lady",
     output_path: Optional[str] = None,
+    speed: float = 1.0,
+    vol: float = 1.0,
+    pitch: int = 0,
+    sample_rate: int = 32000,
+    bitrate: int = 128000,
+    audio_format: str = "mp3",
+    channel: int = 1,
 ) -> bytes:
     """Synthesize text to speech using MiniMax TTS API (SSE stream, hex-encoded audio)."""
     api_key = _require_api_key(api_key)
@@ -193,15 +200,15 @@ def tts_synthesize(
         "stream": True,
         "voice_setting": {
             "voice_id": voice,
-            "speed": 1,
-            "vol": 1,
-            "pitch": 0,
+            "speed": speed,
+            "vol": vol,
+            "pitch": pitch,
         },
         "audio_setting": {
-            "sample_rate": 32000,
-            "bitrate": 128000,
-            "format": "mp3",
-            "channel": 1,
+            "sample_rate": sample_rate,
+            "bitrate": bitrate,
+            "format": audio_format,
+            "channel": channel,
         },
     }
     try:

skills/cli-anything-minimax/SKILL.md

@@ -45,6 +45,13 @@ cli-anything-minimax tts --text "Hello world" --output hello.mp3
 cli-anything-minimax --json chat --prompt "Hello"
 ```
 
+#### TTS options
+
+`--text` (required) / `--model` / `--voice` / `--output` plus voice and audio
+controls: `--speed` (0.5..2.0), `--vol` (0..10), `--pitch` (-12..12),
+`--sample-rate` (8k/16k/22.05k/24k/32k/44.1k), `--bitrate` (32k/64k/128k/256k),
+`--format` (mp3/pcm/flac), `--channel` (1/2). See `tts --help` for defaults.
+
 ## Command Groups
 
 ### Chat