wiz-tts

Wiz TTS

A simple command-line tool for text-to-speech using OpenAI's API, featuring real-time FFT visualization.

Installation

uv tool install -U wiz-tts

# or if you prefer pip
pip install wiz-tts

Usage

After installation, you can run the tool with:

# Recommended: run with uv for best performance
uv run -- wiz-tts "Your text to convert to speech"

# Alternatively, run directly
wiz-tts "Your text to convert to speech"

Or pipe text from another command:

echo "Your text" | uv run -- wiz-tts
cat file.txt | uv run -- wiz-tts

Fun Zsh Function

genz() {
    [[ -z "$1" ]] && echo "Error: URL is required" >&2 && return 1
    local content=$(w3m -dump "$1")
    local persona="gen-z podcaster, cringe but cool and on fire"
    ollama run llama3.2 <<PROMPT | tee /dev/tty | wiz-tts -i "$persona"
Analyze the contents of this webpage:
$content
---
Act as a ${persona}, and compose a succinct monologue.
[no intros or closing, just the meat\!]
PROMPT
}

Example as Mac Shortcut.app Script

exec /opt/homebrew/bin/zsh -i -c "$(cat <<SHELL
INSTRUCTIONS="Voice: Warm, conversational, and authentic, with natural vocal variety and a relaxed pace that invites listeners to settle in for extended listening. Occasional subtle emphasis on key points without sounding rehearsed.

Punctuation: Thoughtful pauses that allow ideas to breathe, varied sentence lengths to maintain interest, and natural breaks that mimic genuine conversation rather than reading.

Delivery: Balanced and rhythmic with a dynamic range that prevents monotony, incorporating strategic shifts in pace and volume to highlight important information or create narrative tension.

Phrasing: Accessible and relatable, using conversational language with occasional colorful expressions or personal anecdotes to build connection. Ideas flow naturally from one to the next with smooth transitions.

Tone: Engaging and inclusive, striking a balance between informative and entertaining, with moments of gentle humor, curiosity, and genuine enthusiasm speaking with a knowledgeable friend.
"
pbpaste | wiz-tts --voice verse --data-dir "${HOME}/Downloads/wiz-tts-data" --instructions "\$INSTRUCTIONS"
SHELL
)"

Options

usage: wiz-tts [-h] [--voice {alloy,echo,fable,onyx,nova,shimmer,coral}] [--instructions INSTRUCTIONS]
               [--model {tts-1,tts-1-hd,gpt-4o-mini-tts}] [--data-dir DATA_DIR] [--split {period,paragraph}]
               [text]

Convert text to speech with visualization

positional arguments:
  text                  Text to convert to speech (default: reads from stdin or uses a sample text)

options:
  -h, --help            show this help message and exit
  --voice {alloy,echo,fable,onyx,nova,shimmer,coral}, -v {alloy,echo,fable,onyx,nova,shimmer,coral}
                        Voice to use for speech (default: coral)
  --instructions INSTRUCTIONS, -i INSTRUCTIONS
                        Instructions for the speech style
  --model {tts-1,tts-1-hd,gpt-4o-mini-tts}, -m {tts-1,tts-1-hd,gpt-4o-mini-tts}
                        TTS model to use (default: tts-1)
  --split {period,paragraph}
                        Split input text by periods or paragraphs and add natural pauses
  --data-dir DATA_DIR, -d DATA_DIR
                        Directory to save audio files and metadata (default: $WIZ_TTS_DATA_DIR if set)
  --bitrate BITRATE, -b BITRATE
                        Audio bitrate for saved files (default: 24k)

Examples

Basic usage:

uv run -- wiz-tts "Hello, world!"

Using stdin:

echo "Hello from stdin" | uv run -- wiz-tts

Using a different voice:

uv run -- wiz-tts --voice nova "Welcome to the future of text to speech!"

Adding speech instructions:

uv run -- wiz-tts --voice shimmer --instructions "Speak slowly and clearly" "This is important information."

Using a different model:

uv run -- wiz-tts --model tts-1-hd "This will be rendered in high definition."

Processing a text file:

cat story.txt | uv run -- wiz-tts --voice echo

Splitting text by sentences with natural pauses:

uv run -- wiz-tts --split period "First sentence. Second sentence. And a third one!"

Splitting text by paragraphs with longer pauses:

cat essay.txt | uv run -- wiz-tts --split paragraph --voice verse

Saving audio to a directory (files are saved as they start generating):

uv run -- wiz-tts "Save this speech to a file" --data-dir ./saved_audio

Using environment variable for audio saving:

# Set the environment variable
export WIZ_TTS_DATA_DIR=./saved_speeches

# Run without --data-dir, audio will be saved to ./saved_speeches
uv run -- wiz-tts "This will be saved using the environment variable"

Saving with custom audio compression:

# Higher bitrate for better quality but larger file size
uv run -- wiz-tts "High quality audio" --data-dir ./saved_audio --bitrate 64k

# Lower bitrate for smaller file size
uv run -- wiz-tts "Compressed audio" --data-dir ./saved_audio --bitrate 16k

Features

• Converts text to speech using OpenAI's TTS API
• Real-time FFT (Fast Fourier Transform) visualization during playback
• Multiple voice options
• Custom speech style instructions
• Reads text from command line arguments or stdin
• Supports multiple TTS models
• Text splitting with natural pauses between segments
• Option to save generated audio as WebM files with metadata and configurable compression
• Saves metadata and audio files immediately as they're being generated (no need to wait for completion)

Requirements

• Python 3.12 or higher
• An OpenAI API key set in your environment variables as OPENAI_API_KEY

Environment Variables

• OPENAI_API_KEY: Your OpenAI API key (required)
• WIZ_TTS_DATA_DIR: Default directory for saving audio files (optional, can be overridden with --data-dir)

License

MIT