feat(tts): add Typecast TTS extension

ai_agents/agents/ten_packages/extension/typecast_tts_python/README.md

@@ -0,0 +1,34 @@
+# Typecast TTS Extension
+
+Typecast text-to-speech extension for TEN Framework.
+
+The extension uses the official `typecast-python` SDK and Typecast's
+`/v1/text-to-speech/stream` endpoint. Typecast streams WAV audio at 32 kHz,
+16-bit, mono. The extension strips the initial WAV header and forwards PCM16
+mono audio to TEN.
+
+## Configuration
+
+```json
+{
+  "params": {
+    "api_key": "<typecast-api-key>",
+    "voice_id": "tc_60e5426de8b95f1d3000d7b5",
+    "model": "ssfm-v30",
+    "output": {
+      "audio_format": "wav"
+    }
+  },
+  "chunk_size": 8192
+}
+```
+
+`output.audio_format` is forced to `wav` because TEN consumes PCM audio frames.
+MP3 streaming would require an additional decoder and is intentionally not used.
+
+## Validation
+
+```bash
+tman -y install --standalone
+./tests/bin/start
+```

ai_agents/agents/ten_packages/extension/typecast_tts_python/__init__.py

@@ -0,0 +1,6 @@
+#
+# This file is part of TEN Framework, an open source project.
+# Licensed under the Apache License, Version 2.0.
+# See the LICENSE file for more information.
+#
+from . import addon

ai_agents/agents/ten_packages/extension/typecast_tts_python/addon.py

@@ -0,0 +1,15 @@
+#
+# This file is part of TEN Framework, an open source project.
+# Licensed under the Apache License, Version 2.0.
+# See the LICENSE file for more information.
+#
+from ten_runtime import Addon, TenEnv, register_addon_as_extension
+
+from .extension import TypecastTTSExtension
+
+
+@register_addon_as_extension("typecast_tts_python")
+class TypecastTTSExtensionAddon(Addon):
+    def on_create_instance(self, ten_env: TenEnv, name: str, context) -> None:
+        ten_env.log_info("TypecastTTSExtensionAddon on_create_instance")
+        ten_env.on_create_instance_done(TypecastTTSExtension(name), context)

ai_agents/agents/ten_packages/extension/typecast_tts_python/config.py

@@ -0,0 +1,71 @@
+#
+# This file is part of TEN Framework, an open source project.
+# Licensed under the Apache License, Version 2.0.
+# See the LICENSE file for more information.
+#
+from pathlib import Path
+from typing import Any
+import copy
+
+from pydantic import Field
+from ten_ai_base import utils
+from ten_ai_base.tts2_http import AsyncTTS2HttpConfig
+
+
+class TypecastTTSConfig(AsyncTTS2HttpConfig):
+    """Typecast TTS config."""
+
+    dump: bool = Field(default=False, description="Typecast TTS dump")
+    dump_path: str = Field(
+        default_factory=lambda: str(
+            Path(__file__).parent / "typecast_tts_in.pcm"
+        ),
+        description="Typecast TTS dump path",
+    )
+    host: str = Field(
+        default="https://api.typecast.ai",
+        description="Typecast API host",
+    )
+    params: dict[str, Any] = Field(
+        default_factory=dict,
+        description="Typecast TTS params",
+    )
+    chunk_size: int = Field(
+        default=8192,
+        ge=1,
+        description="Maximum bytes read from each Typecast streaming chunk",
+    )
+
+    def update_params(self) -> None:
+        if "host" in self.params:
+            self.host = str(
+                self.params.pop("host")  # pylint: disable=no-member
+            ).rstrip("/")
+        else:
+            self.host = self.host.rstrip("/")
+
+        if "model" not in self.params:
+            self.params["model"] = "ssfm-v30"
+
+        output = dict(
+            self.params.get("output") or {}  # pylint: disable=no-member
+        )
+        output["audio_format"] = "wav"
+        self.params["output"] = output
+
+    def validate(self) -> None:
+        if not self.params.get("api_key"):  # pylint: disable=no-member
+            raise ValueError("api_key is required for Typecast TTS")
+        if not self.params.get("voice_id"):  # pylint: disable=no-member
+            raise ValueError("voice_id is required for Typecast TTS")
+        if not self.params.get("model"):  # pylint: disable=no-member
+            raise ValueError("model is required for Typecast TTS")
+
+    def to_str(self, sensitive_handling: bool = True) -> str:
+        if not sensitive_handling:
+            return f"{self}"
+
+        config = copy.deepcopy(self)
+        if config.params and "api_key" in config.params:
+            config.params["api_key"] = utils.encrypt(config.params["api_key"])
+        return f"{config}"

ai_agents/agents/ten_packages/extension/typecast_tts_python/extension.py

@@ -0,0 +1,37 @@
+#
+# This file is part of TEN Framework, an open source project.
+# Licensed under the Apache License, Version 2.0.
+# See the LICENSE file for more information.
+#
+from ten_ai_base.tts2_http import (
+    AsyncTTS2HttpClient,
+    AsyncTTS2HttpConfig,
+    AsyncTTS2HttpExtension,
+)
+from ten_runtime import AsyncTenEnv
+
+from .config import TypecastTTSConfig
+from .typecast_tts import TYPECAST_STREAM_SAMPLE_RATE, TypecastTTSClient
+
+
+class TypecastTTSExtension(AsyncTTS2HttpExtension):
+    """Typecast TTS extension using Typecast streaming TTS."""
+
+    def __init__(self, name: str) -> None:
+        super().__init__(name)
+        self.config: TypecastTTSConfig = None
+        self.client: TypecastTTSClient = None
+
+    async def create_config(self, config_json_str: str) -> AsyncTTS2HttpConfig:
+        return TypecastTTSConfig.model_validate_json(config_json_str)
+
+    async def create_client(
+        self, config: AsyncTTS2HttpConfig, ten_env: AsyncTenEnv
+    ) -> AsyncTTS2HttpClient:
+        return TypecastTTSClient(config=config, ten_env=ten_env)
+
+    def vendor(self) -> str:
+        return "typecast"
+
+    def synthesize_audio_sample_rate(self) -> int:
+        return TYPECAST_STREAM_SAMPLE_RATE

ai_agents/agents/ten_packages/extension/typecast_tts_python/manifest.json

@@ -0,0 +1,83 @@
+{
+  "type": "extension",
+  "name": "typecast_tts_python",
+  "version": "0.1.0",
+  "dependencies": [
+    {
+      "type": "system",
+      "name": "ten_runtime_python",
+      "version": "0.11"
+    },
+    {
+      "type": "system",
+      "name": "ten_ai_base",
+      "version": "0.7"
+    }
+  ],
+  "package": {
+    "include": [
+      "manifest.json",
+      "property.json",
+      "BUILD.gn",
+      "**.tent",
+      "**.py",
+      "README.md",
+      "pyproject.toml",
+      "requirements.txt"
+    ]
+  },
+  "api": {
+    "interface": [
+      {
+        "import_uri": "../../system/ten_ai_base/api/tts-interface.json"
+      }
+    ],
+    "property": {
+      "properties": {
+        "params": {
+          "type": "object",
+          "properties": {
+            "api_key": {
+              "type": "string"
+            },
+            "host": {
+              "type": "string"
+            },
+            "voice_id": {
+              "type": "string"
+            },
+            "model": {
+              "type": "string"
+            },
+            "language": {
+              "type": "string"
+            },
+            "output": {
+              "type": "object",
+              "properties": {
+                "audio_pitch": {
+                  "type": "int64"
+                },
+                "audio_tempo": {
+                  "type": "float64"
+                },
+                "audio_format": {
+                  "type": "string"
+                },
+                "target_lufs": {
+                  "type": "float64"
+                }
+              }
+            },
+            "seed": {
+              "type": "int64"
+            }
+          }
+        },
+        "chunk_size": {
+          "type": "int64"
+        }
+      }
+    }
+  }
+}

ai_agents/agents/ten_packages/extension/typecast_tts_python/pcm.py

@@ -0,0 +1,41 @@
+#
+# This file is part of TEN Framework, an open source project.
+# Licensed under the Apache License, Version 2.0.
+# See the LICENSE file for more information.
+#
+BYTES_PER_SAMPLE = 2
+NUMBER_OF_CHANNELS = 1
+WAV_HEADER_BYTES = 44
+
+
+class StreamingWavToPcm16:
+    """Strip the first WAV header and return PCM16-aligned chunks."""
+
+    def __init__(self) -> None:
+        self._header = bytearray()
+        self._remainder = bytearray()
+        self._header_stripped = False
+
+    def feed(self, chunk: bytes) -> bytes:
+        if not chunk:
+            return b""
+
+        if not self._header_stripped:
+            needed = WAV_HEADER_BYTES - len(self._header)
+            self._header.extend(chunk[:needed])
+            chunk = chunk[needed:]
+            if len(self._header) < WAV_HEADER_BYTES:
+                return b""
+            self._header_stripped = True
+
+        if self._remainder:
+            chunk = bytes(self._remainder) + chunk
+            self._remainder.clear()
+
+        frame_size = BYTES_PER_SAMPLE * NUMBER_OF_CHANNELS
+        left_size = len(chunk) % frame_size
+        if left_size:
+            self._remainder.extend(chunk[-left_size:])
+            chunk = chunk[:-left_size]
+
+        return bytes(chunk)

ai_agents/agents/ten_packages/extension/typecast_tts_python/property.json

@@ -0,0 +1,13 @@
+{
+  "params": {
+    "api_key": "",
+    "voice_id": "tc_60e5426de8b95f1d3000d7b5",
+    "model": "ssfm-v30",
+    "output": {
+      "audio_format": "wav"
+    }
+  },
+  "chunk_size": 8192,
+  "dump": false,
+  "dump_path": "/tmp/typecast_tts_audio_dump"
+}

ai_agents/agents/ten_packages/extension/typecast_tts_python/pyproject.toml

@@ -0,0 +1,7 @@
+[project]
+name = "typecast-tts-python"
+version = "0.1.0"
+requires-python = ">=3.10"
+dependencies = [
+    "typecast-python==0.3.9",
+]

ai_agents/agents/ten_packages/extension/typecast_tts_python/requirements.txt

@@ -0,0 +1 @@
+typecast-python==0.3.9

ai_agents/agents/ten_packages/extension/typecast_tts_python/tests/bin/start

@@ -0,0 +1,9 @@
+#!/bin/bash
+
+set -e
+
+cd "$(dirname "${BASH_SOURCE[0]}")/../.."
+
+export PYTHONPATH=.ten/app:.ten/app/ten_packages/system/ten_runtime_python/lib:.ten/app/ten_packages/system/ten_runtime_python/interface:.ten/app/ten_packages/system/ten_ai_base/interface:$PYTHONPATH
+
+pytest -s tests/ "$@"