dictpress-tts

dictpress-tts is a fast, portable, text-to-speech utility for dictpress. It converts word definitions to audio using cloud-based TTS providers.

---

Integration with dictpress

dictpress-tts connects to the same PostgreSQL database used by DictPress, accessing the entries table to retrieve dictionary words and definitions. It then generates speech audio files for these entries using a configured voice engine, saving them to the specified output directory for easy integration.

---

Features

• Supports Google TTS (other services can be added in the future)
• Handles TTS API rate limiting with configurable options
• Configurable via CLI flags or TOML

---

Installation

Download a prebuilt binary from Releases, or build from source:

make build

Then, either:

Move the binary to a directory in your PATH, such as /usr/local/bin:

mv dictpress-tts /usr/local/bin/

OR

Export the binary's location to your PATH (for the current session):

export PATH="$PATH:$(pwd)"

---

Usage

$ dictpress-tts --help

Usage of dictpress-tts:
  -db-host string
        PostgreSQL host
  -db-name string
        Name of the PostgreSQL database
  -db-pass string
        PostgreSQL password
  -db-port int
        PostgreSQL port
  -db-user string
        PostgreSQL username
  -file string
        Path to dictpress TOML file (default "./config.toml")
  -tts-api-key string
        API key for TTS provider
  -tts-format string
        Audio output format (e.g., mp3, wav) (default "mp3")
  -tts-lang string
        Language code for TTS (e.g., en-US)
  -tts-out-dir string
        Directory to save TTS audio files (default "tts")
  -tts-pitch float
        TTS pitch in dB
  -tts-provider string
        TTS provider (e.g., google) (default "google")
  -tts-rate-limit int
        Max requests per second to the TTS API (default 1000)
  -tts-speed float
        TTS speech rate multiplier (default 1)
  -tts-voice string
        Voice name to use for TTS
  -tts-volume float
        TTS volume gain in dB
  -version
        Print dictpress-tts version
  -workers int
        Number of concurrent TTS processing workers (default 1)

---

Example: Building a voice corpus with Google TTS

This is an example with which Alar's voice corpus can be built. Grab your Google TTS API Key and create a [tts] block in the dictpress TOML file. The language_code and voice_name can be obtained from here.

[tts]
provider = "google"
api_key = "<YOUR-API-KEY-HERE>"
language_code = "kn-IN"
voice_name = "kn-IN-Standard-A"
output_format = "mp3"
out_dir = "tts"
req_per_sec = 10 # requests per second, tune it as per the API rate limits

Make sure your dictpress postgres database is up and running. In case you want to seed the DB with dummy data, you can run the dump shell script provided in the repository.

$ chod +x ./dump # give permission to run as executable
$ ./dump 10000 # optional, dumps 10000 dummy rows into postgres entries table
$ dictpress-tts --file /path/to/config.toml

This will create a tts folder and start writing .mp3 audio files. Depending on the API rate limits and the size of the data, it may take a few minutes to a few hours to finish building the corpus.

---

Running MacOS Binaries

On MacOS, Gatekeeper might prevent you from executing the binary. To prevent Gatekeeper from interfering, com.apple.quarantine attribute needs to be removed.

$ xattr -rd com.apple.quarantine ./dictpress-tts