Description #

Speech synthesis module – text-to-speech solution based on the Doubao voice large model

Key features #

• Integration with the Volcano Engine large model (volc.tts) speech synthesis API
• Supports multiple parameter configurations, such as timbre, emotion, encoding format, speech rate, sampling rate, and more
• Processes multiple text inputs concurrently, outputs Base64-encoded audio and original response
• Suitable for scenarios such as voice broadcasting, virtual human voice generation, and audio content production

Use cases #

• Intelligent customer service, voice assistant content generation
• Multilingual text broadcasting system
• Content platforms, short video dubbing
• Virtual human voice synthesis

Enterprise access instructions #

This capability is currently available only to users who have completed enterprise authentication. To test or for formal integration, please complete the Volcano Engine enterprise authentication process first.

Operator parameters #

Input #

Output #

String array, audio results generated from text;
If recognition fails or the input is an empty string, the corresponding element will be None

Parameters #

If a parameter does not have a default value, it is required

Examples #

The following code demonstrates how to use daft to run the operator and convert text to speech.

from __future__ import annotations

import os

import daft
from daft import col
from daft.las.functions.audio.audio_tts_doubao import AudioTtsDoubao
from daft.las.functions.udf import las_udf

if __name__ == "__main__":
    # Read authentication information for Doubao speech service from environment variables:
    # - OPENSPEECH_APPID: Application ID, used to identify the caller
    # - OPENSPEECH_TOKEN: Access token, used for API authentication
    appid = os.getenv("OPENSPEECH_APPID")
    token = os.getenv("OPENSPEECH_TOKEN")

    if os.getenv("DAFT_RUNNER", "native") == "ray":
        import logging

        import ray

        def configure_logging():
            logging.basicConfig(
                level=logging.INFO,
                format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
                datefmt="%Y-%m-%d %H:%M:%S",
            )
            logging.getLogger("tracing.span").setLevel(logging.WARNING)
            logging.getLogger("daft_io.stats").setLevel(logging.WARNING)
            logging.getLogger("DaftStatisticsManager").setLevel(logging.WARNING)
            logging.getLogger("DaftFlotillaScheduler").setLevel(logging.WARNING)
            logging.getLogger("DaftFlotillaDispatcher").setLevel(logging.WARNING)

        ray.init(dashboard_host="0.0.0.0", runtime_env={"worker_process_setup_hook": configure_logging})
        daft.set_runner_ray()

    daft.set_execution_config(actor_udf_ready_timeout=600)
    daft.set_execution_config(min_cpu_per_task=0)

    samples = {
        "text_input": [
            "",
        ]
    }

    df = daft.from_pydict(samples)
    df = df.with_column(
        "tts_audio",
        las_udf(
            AudioTtsDoubao,
            construct_args={
                "appid": appid,
                "token": token,
                "uid": "test",
                "concurrency": 1,
                "timeout": 60,
            },
            num_cpus=1,
            batch_size=1,
            concurrency=1,
        )(col("text_input")),
    )

    df.show()
    # ╭─────────────────────────────────────────────┬────────────────────────────────╮
    # │ text_input                                  ┆ tts_audio                      │
    # │ ---                                         ┆ ---                            │
    # │ Utf8                                        ┆ Binary                         │
    # ╞═════════════════════════════════════════════╪════════════════════════════════╡
    # │ Welcome to the Doubao Speech Large Model. This is a demonstration. …        ┆ b"\xff\xf3\xe4\xc4\x00\x00\x0… │
    # ╰─────────────────────────────────────────────┴────────────────────────────────╯