Text-to-Speech (TTS) Capability Integration

← Developer Docs · 简体中文

Is Code Required?

Always required. The TTS API has no industry-standard format. Each vendor uses different proprietary API formats, so regardless of whether the platform is OpenAI-compatible for chat, you need to write a Service class.

Note: The TTS service is specified by system-level configuration; users cannot switch between providers. Only one TTS service is active at a time. The system determines the active TTS service via the adi_sys_config table record where name = 'tts_setting'. The value is in JSON format:

{"synthesizer_side": "server", "model_name": "new-model-name", "platform": "newai"}

You can change this via the admin backend, or update the database directly:

UPDATE adi_sys_config SET value = '{"synthesizer_side":"server","model_name":"newai-tts-v1","platform":"newai"}' WHERE name = 'tts_setting';

Prerequisite: Register Platform Name Constant

Add a new platform name constant in AdiConstant.ModelPlatform:

File: adi-common/src/main/java/com/moyz/adi/common/cosntant/AdiConstant.java

public static class ModelPlatform {
    // ... existing constants ...
    public static final String NEWAI = "newai";
}

Step 1: Service Development

Inheritance: CommonModelService → AbstractTtsModelService → Your Service

The base class AbstractTtsModelService is thin (~30 lines), defining only the contract for three abstract methods and proxy settings, so most of the logic must be implemented by subclasses.

Required abstract methods:

Method	Description
start(String jobId, String voice, Consumer<ByteBuffer> onProcess, Consumer<String> onComplete, Consumer<String> onError)	Initialize TTS task, set voice and callbacks
processByStream(String jobId, String partText)	Receive streaming text fragments and send to TTS engine
complete(String jobId)	Finish all text input and wait for audio generation to complete

Complexity Notes

TTS has the highest implementation complexity among all capabilities. It requires handling audio streams, WAV header generation, temporary file management, and other low-level details:

• Streaming API mode (see DashScopeTtsService, ~215 lines): Uses SDK to receive audio frames in real-time, processes PCM data and converts to WAV format, involves audio buffer operations and callback mechanisms
• REST API mode (see SiliconflowTtsService, ~154 lines): Accumulates text first, sends a single request on completion, then processes streaming audio response and writes to file

Step 2: Register in AiModelInitializer

File: adi-common/src/main/java/com/moyz/adi/common/service/AiModelInitializer.java

Add in the initTtsModelServiceList() method:

private synchronized void initTtsModelServiceList(Map<String, ModelPlatform> nameToPlatform) {
    // ... existing platforms ...

    // Add new
    initTtsModelService(AdiConstant.ModelPlatform.NEWAI,
        model -> new NewAiTtsService(model, nameToPlatform.get(AdiConstant.ModelPlatform.NEWAI)));
}

---

Previous: Text-to-Image Integration · Next: ASR Integration