Configuration (config.json)

slidemovie allows you to customize video resolution, audio quality, and TTS settings via JSON configuration files.

Configuration Loading Order

The program loads settings in the following order (later sources override earlier ones):

  1. Built-in Defaults: Hardcoded in the program.
  2. User Config: ~/.config/slidemovie/config.json (Home directory).
    • Note: If this file does not exist, it will be automatically created with default values on the first run.
  3. Local Config: ./config.json (Current working directory).
  4. CLI Arguments: Command-line flags (highest priority for specific settings).

Configuration Options

Here is a complete list of available keys in config.json.

Text-to-Speech (TTS)

See multiai-tts for supported providers and models.

Key Type Default Description
tts_provider string "google" The AI provider (google, openai, azure, or voicevox).
tts_model string "gemini-3.1-flash-tts-preview" The specific model name. Required for Google and OpenAI; ignored for Azure and VOICEVOX.
tts_voice string "sadaltager" The voice ID (e.g., alloy for OpenAI, en-US-AvaMultilingualNeural for Azure and English). For VOICEVOX, an integer speaker style ID (e.g., "3").
tts_use_prompt bool true Whether to send a system prompt to the TTS API. Set true for google and false for OpenAI, Azure and VOICEVOX.
prompt string "Please speak..." The system instruction for the TTS engine.
prompt_separator string "" Separator inserted between the style prompt (prompt + per-slide additional_prompt) and the spoken text. Empty disables it (default). Useful with long prompts to clearly mark where the script begins — e.g. "\n\nScript:\n". See Custom Prompt per Slide.
chunk_size int / null null Max characters per chunk for long narration. null disables splitting (default behavior). Measured against the spoken text only, not the prompt. See Long narration.
split_chars string "。．.!！?？\n" Candidate characters where the text may be split into chunks.
chunk_overflow string "extend" Behavior when no split candidate is found within chunk_size: "extend" reads on to the next candidate; "error" stops with an error.
tts_voicevox_url string / null null VOICEVOX engine URL. null uses the default (http://127.0.0.1:50021). Only used when tts_provider is voicevox.

VOICEVOX: VOICEVOX runs as a local engine and must already be running before you build (default http://127.0.0.1:50021). No API key is required, tts_model is ignored, and tts_voice must be an integer speaker style ID. Set tts_use_prompt to false so the style prompt is not read aloud.

Video Format

Key Type Default Description
screen_size [int, int] [1280, 720] Resolution [width, height]. Slide images are normalized to this size.
image_pad_color string "white" Padding color used when a slide’s aspect ratio differs from screen_size. The image is scaled to fit and centered, with even letterbox/pillarbox padding of this color. Any ImageMagick color name/value is accepted (e.g. "black").
video_fps int 30 Frames per second.
video_codec string "libx264" Video encoding codec.
video_pix_fmt string "yuv420p" Pixel format (compatibility).

Audio Format

Key Type Default Description
audio_codec string "aac" Audio encoding codec.
sample_rate int 44100 Sample rate in Hz.
audio_bitrate string "192k" Audio quality bitrate.
audio_channels int 2 1 (Mono) or 2 (Stereo).

General / Processing

Key Type Default Description
silence_sec float 2.5 Silence duration added before each slide speaks.
max_retry int 2 Number of retries if TTS API fails.
ffmpeg_loglevel string "error" Log verbosity for FFmpeg processes.
show_skip bool false If true, logs “skipped” tasks (unchanged files) to the console. Can be enabled via --debug.
output_root string null Custom root directory for video output. Can be overridden by the -o CLI option.
output_filename string null Output video filename (without extension). Defaults to project ID. Can be overridden by the -f CLI option.

Example config.json

To create a Full HD (1080p) video using OpenAI, save this as config.json in your project folder:

{
    "tts_provider": "openai",
    "tts_model": "gpt-4o-mini-tts",
    "tts_voice": "alloy",
    "tts_use_prompt": false,
    "screen_size": [1920, 1080]
}

To use VOICEVOX (start the local engine first):

{
    "tts_provider": "voicevox",
    "tts_voice": "3",
    "tts_use_prompt": false
}

Warning: Once you start building a project, the screen_size and FPS are recorded in the status file (status.json). If you change these settings halfway through, slidemovie will detect an inconsistency and abort to prevent video corruption. To apply new video settings, you may need to delete the files in movie/ directory to rebuild from scratch.