Skip to content

Full Settings Rundown (WIP)

Jump to bottom
Beangate edited this page Jul 8, 2025 · 5 revisions

GameSentenceMiner Configuration Guide

This document details every setting available in the configuration window to help you tailor the application to your specific sentence mining workflow.

Legend for Settings

Throughout the configuration window, you'll notice that some setting labels are color-coded. Here's what they mean:

  • Orange Text: Important / Required Setting. These settings are crucial for the core functionality of GSM. Incorrect or missing values for these settings will likely cause the application to fail or not work as expected.
  • Red Text: Advanced / Potentially Risky Setting. These are advanced options that can significantly alter behavior. They may have unintended side effects or require a deeper understanding of the underlying tools (like FFmpeg). Use them with caution.
  • Green Text: Recommended / Quality-of-Life Setting. These settings are optional but highly recommended for a better, more organized, or more efficient user experience.

Table of Contents

  1. General Tab
  2. Paths Tab
  3. Anki Tab
  4. VAD (Voice Activity Detection) Tab
  5. Features Tab
  6. Screenshot Tab
  7. Audio Tab
  8. OBS Tab
  9. Profiles Tab
  10. AI Tab
  11. Advanced Tab

General Tab

This tab contains the most fundamental settings for how GSM receives input and general application behavior.

Websocket Enabled

Tooltip: Enable or disable WebSocket communication. Enabling this will disable the clipboard monitor.

Default Value: True

Details: This is one of the two primary methods for getting text into GSM. When enabled, GSM opens a WebSocket server that can receive text from other applications, most commonly a browser-based texthooker like the one built into GSM. This method is generally faster and more reliable than clipboard monitoring. This is the recommended method for input, and clipboard monitoring is disabled by default when this is on.

Clipboard Enabled

Tooltip: Enable or disable Clipboard monitoring.

Default Value: True

Details: This is the second primary method for getting text into GSM. When enabled, GSM will constantly monitor your system's clipboard. Any time new text is copied, GSM will process it as a new line. Use this if your texthooking method doesn't support WebSockets. It's a reliable fallback but can be accidentally triggered.

Allow Both Simultaneously

Tooltip: Enable to allow GSM to accept both clipboard and websocket input at the same time.

Default Value: False

Details: By default, GSM will prioritize one input method. This advanced setting allows both the WebSocket server and the clipboard monitor to be active at once. This can lead to duplicate entries if you are not careful (e.g., if your texthooker sends via WebSocket and copies to the clipboard). Only enable this for specific workflows that require simultaneous input from different sources.

Websocket URI(s)

Tooltip: WebSocket URI for connecting. Allows Comma Separated Values for Connecting Multiple.

Default Value: localhost:6677,localhost:9001,localhost:2333 (Textractor, Agent, Luna)

Details: Specifies the address(es) and port(s) for the WebSocket server(s) that GSM will listen to. You can provide a comma-separated list to listen to multiple sources (e.g., a texthooker and an OCR tool) simultaneously. The default value is set to listen on several common ports for texthooking tools.

Texthook Replacement Regex

Tooltip: Regex to run replacement on texthook input, set this to the same as what you may have in your texthook page.

Default Value: "" (empty string)

Details: Allows you to specify a Regular Expression (Regex) to automatically clean up incoming text before GSM processes it. This is useful for removing artifacts like control codes or speaker names. This should mirror any replacement rules you have set up in your texthooker.

Open Config on Startup

Tooltip: Whether to open config when the script starts.

Default Value: False

Details: A convenience setting. If checked, this configuration window will automatically appear every time you launch GameSentenceMiner.

Open GSM Texthooker on Startup

Tooltip: Whether to open Texthooking page when the script starts.

Default Value: True

Details: If checked, your default web browser will automatically open the built-in GSM texthooker page on startup.

GSM Texthooker Port

Tooltip: Port for the Texthooker to run on. Only change if you know what you are doing.

Default Value: 55000

Details: Specifies the network port for the built-in texthooker web page. You should only change this if the default port is already in use by another application on your system.

Paths Tab

This tab is for configuring all the essential file and folder locations.

Folder to Watch

Tooltip: Path where the OBS Replays will be saved.

Default Value: ~/Videos/GSM

Details: This is a critical path setting. It must point to the exact directory where OBS is configured to save its Replay Buffer files. The ~ symbol automatically expands to your user's home directory (e.g., C:\Users\YourName on Windows).

Audio Destination

Tooltip: Path where the cut Audio will be saved.

Default Value: ~/Videos/GSM/Audio/

Details: This is the folder where the final, trimmed audio clips will be saved.

Screenshot Destination

Tooltip: Path where the Screenshot will be saved.

Default Value: ~/Videos/GSM/SS/

Details: Similar to the Audio Destination, this is where the final screenshot images will be saved.

Remove Video

Tooltip: Remove video from the output.

Default Value: True

Details: If enabled, GSM will delete the original OBS replay video file after it has successfully processed it. This is useful for saving disk space.

Remove Audio

Tooltip: Remove audio from the output.

Default Value: False

Details: If enabled, the temporary, untrimmed audio file created during processing will be deleted. This is generally recommended to keep your processing folders clean.

Remove Screenshot

Tooltip: Remove screenshots after processing.

Default Value: False

Details: If enabled, GSM will delete the intermediate screenshot file after it's been moved to the final destination and added to Anki.

Anki Tab

This tab controls all interactions with Anki, including which fields to update and how to tag your cards.

Update Anki

Tooltip: Automatically update Anki with new data.

Default Value: True

Details: The master switch for all Anki integration. If disabled, GSM will process media but will not attempt to connect to Anki or update any cards.

Anki URL

Tooltip: The URL to connect to your Anki instance.

Default Value: http://127.0.0.1:8765

Details: The address for the AnkiConnect add-on, which allows GSM to communicate with Anki. The default value should work for most users. You must have the AnkiConnect add-on installed and Anki running.

Field Mappings

Details: These fields are the core of the Anki integration. You must enter the exact names (case-sensitive) of the fields from your Anki Note Type that you want GSM to populate.

  • Sentence Field: Default: "Sentence"
  • Sentence Audio Field: Default: "SentenceAudio"
  • Picture Field: Default: "Picture"
  • Word Field: Default: "Expression"
  • Previous Sentence Field: Default: "" (empty)
  • Previous Image Field: Default: "" (empty)

Add Tags

Tooltip: Comma-separated custom tags for the Anki cards.

Default Value: GSM

Details: Tags listed here (comma-separated) will be added to every Anki card that GSM updates.

Tags to work on

Tooltip: Comma-separated Tags, script will only do 1-click on cards with these tags...

Default Value: (empty)

Details: This setting acts as a filter for "Full Auto Mode". If empty, GSM targets the most recently created Anki card. If filled (e.g., gsm-mine), GSM will only operate on cards that already have that specific tag.

Add Game as Tag & Game Parent Tag

Tooltip (Add Game): Include a tag for the game on the Anki card. Tooltip (Parent Tag): Parent tag for the Game Tag...

Default Value (Add Game): True Default Value (Parent Tag): Game

Details: These settings work together to automatically organize cards by game. The "Current Game" is determined by your active OBS Scene. For example, if Parent Tag is Game and the current scene is FinalFantasyX, the resulting Anki tag will be Game::FinalFantasyX, creating a nested tag in Anki.

Overwrite Audio / Overwrite Picture

Tooltip: Overwrite existing audio/pictures in Anki cards.

Default Value (Audio): False Default Value (Picture): True

Details: If enabled, GSM will replace any existing content in the respective field. If disabled, it will not add new media if the field is already filled.

Multi-line Mining Overwrite Sentence

Tooltip: When using Multi-line Mining, overwrite the sentence with a concatenation of the lines selected.

Default Value: True

Details: When using the multi-line mining feature, this setting, if enabled, will join the selected lines and overwrite the content of the Sentence Field in Anki.

VAD (Voice Activity Detection) Tab

This tab contains settings for automatically trimming audio clips to isolate spoken dialogue.

Voice Detection Postprocessing

Tooltip: Enable post-processing of audio to trim just the voiceline.

Default Value: True

Details: This is the master switch for VAD. If enabled, it will use the models and settings on this tab to intelligently find and trim the audio. If disabled, GSM takes a raw audio slice based on offsets.

Language

Tooltip: Select the language for VAD.

Default Value: ja (Japanese)

Details: This hint helps the VAD models (especially Whisper) more accurately detect speech in the target language.

Whisper Model

Tooltip: Select the Whisper model size for VAD.

Default Value: base

Details: Determines the size and accuracy of the local Whisper model. Larger models (small, medium) are more accurate but slower and require more resources. tiny and base are faster.

Select VAD Model

Tooltip: Select which VAD model to use.

Default Value: whisper

Details: The primary engine for voice detection. whisper is accurate but can be slow. silero and vosk are faster local alternatives.

Backup VAD Model

Tooltip: Select which model to use as a backup if no audio is found.

Default Value: silero

Details: If the primary VAD model fails to find speech, GSM can try again with this backup model. The default setup uses the accurate whisper model first, then falls back to the fast silero model.

Add Audio on No Results

Tooltip: Add audio even if no results are found by VAD.

Default Value: False

Details: If enabled, and if both primary and backup VAD models find no speech, GSM will save the original, untrimmed audio clip instead of failing.

Audio End Offset

Tooltip: Offset in seconds from end of the video to extract.

Default Value: 0.5

Details: This setting defines the length of the initial audio clip that is fed into the VAD model. It's a critical tuning parameter. It is added to the duration of the video replay segment. A value of 0.5 means the clip will be half a second longer than the detected speech segments.

Trim Beginning / Beginning Offset

Tooltip: Beginning offset after VAD Trim...

Default Value (Trim Beginning): False Default Value (Beginning Offset): -0.25

Details: After VAD finds the start of the speech, this lets you fine-tune it. A small negative offset (like the default -0.25s) is useful to add a tiny bit of silence before the speech begins, ensuring the sound isn't clipped. This only applies if "Trim Beginning" is checked.

Cut and Splice Segments / Padding

Tooltip: Cut Detected Voice Segments and Paste them back together...

Default Value (Cut and Splice): False Default Value (Padding): 0.1

Details: An advanced feature for audio clips with multiple pauses. When enabled, VAD finds all distinct speech segments, cuts them out, and stitches them together into one file, with a silence of Padding seconds between each segment.

Features Tab

This tab controls various automated behaviors and quality-of-life features.

Full Auto Mode

Tooltip: Yomitan 1-click anki card creation.

Default Value: True

Details: Enables the "one-click" workflow. In this mode, GSM listens for new card creations in Anki (e.g., from Yomitan). When it detects one, it automatically grabs the most recent replay and adds the corresponding audio/screenshot to that new card.

Notify on Update

Tooltip: Notify the user when an update occurs.

Default Value: True

Details: If enabled, GSM shows a system notification pop-up whenever it successfully updates an Anki card.

Open Anki Edit / Open Anki Note in Browser

Tooltip: Automatically open Anki for editing... / Open Anki note in browser...

Default Value (Edit): False Default Value (Browser): True

Details: Provides immediate feedback. After updating a card, GSM can open the Anki card editor window, or open the Anki browser focused on the card that was just updated.

Browser Query

Tooltip: Query to use when opening Anki notes in the browser. Ex: 'Added:1'

Default Value: "" (empty string)

Details: When "Open Anki Note in Browser" is enabled, this Anki search query determines which cards are shown. An empty query will default to showing the single card that was just modified. A query like added:1 would show all cards added in the last day.

Backfill Audio

Tooltip: Fill in audio data for existing entries.

Default Value: False

Details: A special mode to add audio/screenshots to existing cards. When you trigger a mine, GSM will search for an Anki card with that exact sentence and add the media to it. Warning: This cannot be used at the same time as Full Auto Mode. PROBABLY DOESN'T WORK. IM NOT SURE.

Screenshot Tab

This tab controls everything related to capturing, processing, and timing screenshots.

Enabled

Tooltip: Enable or disable screenshot processing.

Default Value: True

Details: The master switch for all screenshot functionality.

Width, Height, Quality, Extension

Details: These settings control the final output of the image.

  • Width: Default: 0 (0 means do not resize width)
  • Height: Default: 0 (0 means do not resize height)
  • Quality: Default: 85 (For lossy formats like jpeg and webp, controls compression from 0-100)
  • Extension: Default: "webp" (webp or avif are recommended for good quality/size balance)

FFmpeg Reencode Options

Tooltip: Custom FFmpeg options for re-encoding screenshots.

Default Value: "" (empty string)

Details: For advanced users. This field allows you to provide your own custom FFmpeg command-line arguments for image conversion, overriding the basic settings above.

Screenshot Timing & Screenshot Offset

Tooltip (Timing): Select when to take the screenshot relative to the line: beginning, middle, or end. Tooltip (Offset): Time in seconds to offset the screenshot...

Default Value (Timing): beginning Default Value (Offset): 1.0

Details: These settings are crucial for getting the perfect screenshot. The final timestamp is calculated based on the timing anchor plus the offset.

  • Example: With default settings, the screenshot is taken 1.0 second after the beginning of the line. To take a screenshot 0.5 seconds after a voice line finishes, you would set Timing to end and Offset to 0.5.

Use Screenshot Selector for every card

Tooltip: Enable to use the screenshot selector to choose the screenshot point on every card.

Default Value: False

Details: If enabled, a manual frame-by-frame screenshot selection tool will pop up for every single card, allowing you to choose the exact frame you want.

Screenshot Hotkey Updates Anki

Tooltip: Enable to allow Screenshot hotkey/button to update the latest anki card.

Default Value: False

Details: If enabled, using the "Take Screenshot" hotkey will not only save a screenshot but also attempt to add it to the most recently updated Anki card.

Audio Tab

This tab controls the specifics of audio extraction and processing.

Enabled

Tooltip: Enable or disable audio processing.

Default Value: True

Details: The master switch for all audio functionality.

Audio Extension

Tooltip: File extension for audio files.

Default Value: opus

Details: The audio format for the final clip. opus is highly recommended for its excellent compression and quality.

Audio Extraction Beginning Offset

Tooltip: Offset in seconds from beginning of the video to extract (Should Usually be negative or 0).

Default Value: -0.5

Details: This critical setting adjusts the start time of the initial audio slice relative to when the line was registered. Since the replay buffer captures video before you trigger it, this value is almost always negative. It determines how much time before the registered line is included in the audio clip, ensuring the beginning of the speech is not cut off.

Audio Extraction End Offset

Tooltip: Offset in seconds to trim from the end before VAD processing starts.

Default Value: 0.0

Details: An advanced parameter that shortens the initial audio clip from the end before it's sent to VAD. The default of 0.0 does nothing. A positive value can be useful for getting extra audio audio after the speech ends, while a negative value can help remove unwanted noise or echoes at the end.

FFmpeg Reencode Options & FFmpeg Preset Options

Tooltip: Custom FFmpeg options... / Select a preset FFmpeg option...

Default Value: -c:a libopus -f opus -af "afade=t=in:d=0.10" (on Windows)

Details: These control how the final audio clip is encoded.

  • Preset Options: A dropdown of tested FFmpeg commands for common tasks like normalizing volume (loudnorm) or adding a fade-in to prevent clipping (the default behavior). This is the recommended way to set these options.
  • Reencode Options: If you select "Custom" from the presets, you can write your own FFmpeg command here for full control.

External Audio Editing Tool

Tooltip: Path to External tool that opens the audio up for manual trimming...

Default Value (Path): "" (empty string) Default Value (Enabled): True

Details: Allows integration of a third-party audio editor. If a path is provided and this is enabled, GSM will open the trimmed audio file in the specified program (e.g., OcenAudio, Audacity) for manual review before saving to Anki. The "Install Ocenaudio" button can set this up automatically.

Anki Media Collection

Tooltip: Path of the Anki Media Collection, used for external Trimming tool.

Default Value: "" (empty string)

Details: This should be the full path to your collection.media folder inside your Anki profile directory. It's used by the External Audio Editing Tool feature to save the file in the correct place. The "Get Anki Media Collection" button can find and set this path for you.

OBS Tab

This tab controls the integration with OBS (Open Broadcaster Software).

Enabled, Host, Port, Password

Details: These settings are for connecting to the OBS WebSocket plugin, which must be enabled in OBS (Tools > WebSocket Server Settings). The values here must match your OBS configuration.

  • Enabled: Default: True
  • Host: Default: "127.0.0.1"
  • Port: Default: 7274
  • Password: Default: "your_password" (You must change this to your actual OBS WebSocket password!)

Open OBS / Close OBS

Tooltip: Open OBS when GSM starts / Close OBS when GSM closes.

Default Value (Open): True Default Value (Close): True

Details: Convenience settings to manage the OBS application alongside GSM.

Get Game From Scene Name

Tooltip: Changes Current Game to Scene Name

Default Value: True

Details: A powerful automation feature. If enabled, when you switch scenes in OBS, GSM will check for a Profile with a matching name and automatically activate it.

Minimum Replay Size (KB)

Tooltip: Minimum Replay Size for OBS Replays in KB. If Replay is Under this, Audio/Screenshot Will not be grabbed.

Default Value: 0

Details: A filter to prevent processing of empty or corrupted replay files. If a new replay video is smaller than this size in kilobytes, GSM will ignore it. A value of 0 disables this check.

Profiles Tab

This tab allows you to manage multiple configurations for different games or use cases.

Select Profile

Tooltip: Select a profile to load its settings. Details: The dropdown shows all your saved profiles. Selecting a profile immediately loads all its associated settings across all tabs. The Default profile cannot be deleted.

Add / Copy / Delete Profile

Tooltip: Standard profile management buttons. Details:

  • Add Profile: Creates a new profile based on the default settings.
  • Copy Profile: Creates a duplicate of the currently selected profile.
  • Delete Profile: Deletes the currently selected profile (except for Default).

OBS Scene (Auto Switch Profile)

Tooltip: Select an OBS scene to associate with this profile. (Optional) Details: This is the core of the auto-profile switching feature. Here, you can link one or more OBS scenes to the current profile. When OBS switches to any of the selected scenes, GSM will automatically activate this profile. You can select multiple scenes by holding Ctrl or Shift. I've run into issues with this feature, so use it with caution.

AI Tab

This tab configures generative AI features, such as automatic translations or context generation.

Enabled

Tooltip: Enable or disable AI integration.

Default Value: False

Details: Master switch for AI features. Requires an internet connection and a valid API key.

Provider, Model, and API Key

Details: Choose your AI service, the specific model, and enter your secret API key.

  • Provider: Default: 'gemini'
  • Gemini Model: Default: 'gemini-2.0-flash'
  • Groq Model: Default: 'meta-llama/llama-4-scout-17b-16e-instruct'
  • Gemini/Groq API Key: Default: "" (empty string)

Anki Field

Tooltip: Field in Anki for AI-generated content.

Default Value: "" (empty string)

Details: The name of the Anki field where the output from the AI will be placed (e.g., a Translation or AI_Context field).

Use Canned... / Custom Prompt

Details: This controls the instructions given to the AI.

  • Use Canned Translation Prompt: Default: True
  • Use Canned Context Prompt: Default: False
  • Custom Prompt: Default: "" (empty string)

Advanced Tab

This tab contains miscellaneous settings for advanced users or niche workflows.

Audio Player Path / Video Player Path

Tooltip: Path to the audio/video player executable.

Default Value: "" (empty string)

Details: Specify a media player (like MPC-HC or VLC) to quickly preview your content using the hotkey below. The video player option will even seek to the correct timestamp in the original replay file.

Hotkeys

Details: Configure keyboard shortcuts for various actions.

  • Play Latest Video/Audio Hotkey: Default: 'f7'
  • Take Screenshot Hotkey: Default: 'f6'
  • Reset Line Hotkey: Default: 'f5'

Multi-line Mining Settings

Details: These control the behavior of the multi-line mining feature.

  • Multi-line Line-Break: Default: '<br>' (The character(s) to insert between joined sentences).
  • Multi-Line Sentence Storage Field: Default: "" (A temporary field in Anki to hold the combined text, if needed by your workflow).

Websocket Ports

Details: Configure ports for advanced communication features.

  • OCR WebSocket Port: Default: 9002 (Port for receiving text from external OCR tools).
  • Texthooker Communication WebSocket Port: Default: 55001
  • Plaintext Websocket Export Port: Default: 55002

Polling Rate

Tooltip: Rate at which Anki will check for updates (in milliseconds).

Default Value: 200

Details: In "Full Auto Mode," this determines how frequently GSM asks Anki if a new card has been created. Lower value means faster response but higher resource usage. The default is generally fine. This setting is located on the Anki tab in the UI.

Clone this wiki locally