Verity Mod Java Edition Setup Guide

Recommended Java path

Start with stable Verity JE 5.7.4 for Minecraft 1.20.1 with Forge if you want the release rather than a test build. The box above shows the current stable and test builds. Check the maintainer’s file listing.

What you need	Choose	Before installing
The current release, with the stable setup below	5.7.4, Forge 1.20.1	Use its own profile and required dependencies
Features or add-ons requiring Verity 6.x	6.0.0-beta.9, Forge 1.20.1	Keep a separate test profile; use the 6.0.0+ tutorial
An existing NeoForge 1.21.1 installation	Legacy 3.4.1	Do not use the Forge configuration instructions below

The stable label does not guarantee that every cloud model remains available to every account. If installation succeeds but Groq rejects a model, check that specific error before replacing your key or world. The Groq configuration guide explains the distinction between Verity’s model preset and Groq’s current model availability.

Match the tutorial to the release

• 5.7.x: the earlier Java setup video and Ollama walkthrough titled “up to 5.7.3” are older references. The 5.7.4 changelog records a missing Groq model fix, not a new setup tutorial.
• 6.0.0 beta: CurseForge’s Setup link now opens Full Verity JE Mod Setup | 6.0.0+. The beta.9 changelog specifically records an AI config rework. A field shown in an earlier video may no longer match beta.9.
• Legacy NeoForge: choose the 3.4.1 file under Minecraft 1.21.1 in the maintainer’s file listing. A tutorial for Forge 1.20.1 does not establish which configuration fields exist in this older release.

Installing for friends? Complete the single-player test first, then follow the Verity multiplayer and server checklist. CurseForge’s Client & Server label does not by itself confirm compatibility with a particular hosting service.

What your computer needs

Verity’s own requirements depend heavily on the AI path you choose, so split the question in two.

For Minecraft and Forge alone, the baseline is the same as any modest Forge 1.20.1 profile: a 64-bit operating system, Java 17 or newer, and enough free memory to give the game a few gigabytes. The official launcher and the CurseForge app bundle a suitable Java runtime; problems usually begin when a player has manually pointed a profile at an old Java 8 installation left over from earlier modding.

For the AI path, the demands differ sharply:

AI path	Where the model runs	What your PC must provide
Groq (cloud)	GroqCloud servers	A stable internet connection; very little extra CPU, GPU, or RAM
Ollama (local)	Your own computer	Memory and processing power for the model, alongside Minecraft itself

The default local model in 5.7.4, qwen2.5:1.5b, is a small model chosen so that ordinary machines can run it. Larger models may give richer dialogue but compete with Minecraft for memory; on a laptop, the combination can produce slow replies or stuttering frames. If you are unsure, start with Groq to confirm the mod works, then try local inference as a second step. The Ollama guide covers model size choices in more detail.

What is inside the jar

We opened the 5.7.4 and 6.0.0-beta.9 jars to check the facts tutorials often get wrong:

• Required mods, as written in the jar: Forge 47 or newer, Minecraft 1.20.1, GeckoLib 4.4+, and YetAnotherConfigLib (YACL) 3.6+. YACL is only needed on the client; a dedicated server does not need it.
• Why the file is about 250 MB. Most of it is an offline voice (Piper, about 67 MB) and an offline English speech-recognition model (about 62 MB), plus native libraries for Windows, macOS, and Linux. A download much smaller than that is not the full mod.
• Where settings are saved. The in-game Mods > Verity > Config screen writes Forge config files in the instance’s config folder (verity-client.toml and verity-common.toml for 5.7.4). Editing them while the game is closed has the same effect; never post them, because the API key is stored there in plain text.

Voice options built into 5.7.4

5.7.4 can speak and listen without any cloud service:

Setting	What it does	Default
Use Offline AI Voice	Uses the built-in voice, which runs on your CPU	On
Use Native TTS	Uses your operating system’s narrator instead	Off
Built-in speech recognition	Recognises English speech offline; the setting itself warns it “requires beefy PC”	Off
Use Kokoro / Local Whisper URL	Points voice output or recognition at your own Kokoro (127.0.0.1:8880) or Whisper (127.0.0.1:9000) server	Off

With the defaults, Verity’s voice comes from your own PC, while your microphone input is sent to Groq for transcription. If voice is the only thing that fails, the conversation guide has the step-by-step checks.

Prepare a clean profile

Create a separate launcher profile for the exact Minecraft version. Install the matching loader into that profile: Forge for 1.20.1 or NeoForge for the legacy 1.21.1 branch. Launch that empty loader profile once before adding Verity. A successful empty launch proves that Java, Minecraft, and the loader can start together.

Keep the first test profile small. Do not add a large modpack, shaders, performance patches, and Verity all at once. Every extra mod expands the number of possible conflicts and makes logs harder to read. After Verity works by itself, add other mods in small groups and test between each group.

Back up existing worlds before opening them with a new loader or mod set. A separate test world is preferable because a failed experiment should not become a world-recovery problem.

Launcher-specific notes

The steps are the same everywhere, but the buttons are not:

• CurseForge app: choose Create Custom Profile, select Minecraft 1.20.1 and Forge, then use the profile’s Add Content search to find Verity JE by VarmiteYT. The app places the file in the correct instance and can resolve listed dependencies. Use Open Folder on the profile to inspect the mods directory when diagnosing.
• Official Minecraft Launcher: run the Forge 1.20.1 installer, choose Install client, and launch the new Forge profile once. The installer creates a profile that uses the shared .minecraft folder by default; set a separate Game directory in the profile settings if you want Verity isolated from other mod setups.
• Prism Launcher and similar tools: create an instance for 1.20.1, add Forge under the instance’s version settings, and add the jar through the instance’s mod list. Each instance already has its own folder, which makes clean testing easy.

Whichever launcher you use, the most common mistake is installing into one profile and launching another. If Verity does not appear in Minecraft’s Mods list, check which instance you started before anything else.

Memory allocation

Many launchers default to 2 GB for Minecraft. That is often enough for a small Forge profile, but tight once you add other mods. Raising the allocation to roughly 4 GB is a reasonable first step for a light profile; going far beyond what the profile needs does not make Verity reply faster and can leave less memory for a local Ollama model. If you run Ollama on the same machine, leave headroom for it rather than assigning almost all system memory to Minecraft.

Install from the maintained project

Begin with the download guide and open Verity JE by VarmiteYT, Project ID 1591438. For the stable route, select the 5.7.4 file details: filename verity-5.7.4.jar, Minecraft 1.20.1, loader Forge, release type R. These links open the maintainer’s pages, not a mirrored package.

Open the Verity JE CurseForge project and confirm these four details:

1. Project title: Verity JE.
2. Project author: VarmiteYT.
3. Project ID: 1591438.
4. File target: your exact Minecraft version and loader.

Use the dependencies listed for the specific release at the time you download. Dependency requirements can change, so this guide does not freeze an old list copied from a video. Place the selected project file and required dependencies in the clean profile’s mods directory, not in a different launcher instance that happens to have a similar name.

Launch once and stop at the title screen. If Minecraft fails before a world opens, the problem is not the AI setup: check the Java runtime, loader, missing dependencies, duplicate mods, or incompatible mods. Read the first meaningful error in the latest log instead of focusing only on the final “exit code” line.

Reading the launch result

• Title screen appears and Mods lists Verity: installation is complete; continue to the AI configuration.
• Forge reports missing or mismatched mods: the message names the dependency and the version it expected. Download that exact dependency for 1.20.1 Forge.
• Crash before the Mojang screen: usually the Java runtime or a loader mismatch, not Verity. Confirm Java 17+ and that the profile really is Forge 1.20.1.
• Two Verity jars in the folder: remove the older one. Duplicate copies of the same mod are a frequent cause of confusing startup errors after an update.

Choose one AI path

Both routes start and end in Minecraft; they differ in where the reply is written and which programs must be running.

1. Minecraft + Verity JEYou type in chat
2. HTTPS, with your Groq key
3. Groq (cloud)Writes the reply; needs internet; rate limits apply

1. Minecraft + Verity JEUse Ollama turned on
2. 127.0.0.1:4000
3. LiteLLM gatewayTerminal must stay open
4. 127.0.0.1:11434
5. Ollama + modelWrites the reply on your PC

Yellow border: leaves your computerGreen border: stays on your computer

Verity JE supports a cloud path using Groq and a local path using Ollama. They solve the same core need in different ways.

Groq is usually simpler when your computer should not run a local language model. It requires a GroqCloud account, an active API key, and network access from the game. In 5.7.4, open Mods > Verity > Config from Minecraft’s title screen. The screen is titled Verity Configuration. Under AI Settings, set AI Provider to GROQ, paste the secret into API Key, and leave AI Intelligence Level at FAST for the first test. Under Local LLM, keep Use Ollama off: enabling it overrides the cloud-provider choice. Apply the changes and reopen Config to check them without exposing the key.

Ollama keeps model inference on your own machine, but stable 5.7.4 uses a LiteLLM gateway between Verity and Ollama. In the same config screen, Local LLM contains Use Ollama, LiteLLM URL (default http://127.0.0.1:4000/v1/), and Ollama AI Model (default ollama/qwen2.5:1.5b). The gateway prefix ollama/ is intentional; Ollama itself lists that model as qwen2.5:1.5b. Follow the service and gateway checks before enabling this option.

These labels and defaults were checked in the maintainer’s published 5.7.4 artifact. The 6.0.0 betas have a different configuration flow.

Choose one provider for the first successful conversation. Configuring both at once makes it unclear which provider produced an error. Follow the Groq setup or Ollama setup guide, then return here.

Groq or Ollama: a quick decision

Your situation	Better first choice	Why
Laptop or older PC	Groq	No local model competing with Minecraft for memory
Unreliable or metered internet	Ollama	Conversations do not depend on a remote service
You do not want an online account	Ollama	Inference stays on your machine
You want the fewest setup steps	Groq	One key instead of Ollama plus a LiteLLM gateway
Long sessions where service limits matter	Ollama	No cloud rate limits, only your hardware limits

You can switch later. Keeping the working provider’s settings written down makes it easy to return if the second path does not work.

First world test

Create a new world with default rules unless the project instructions require a specific setting. Confirm that the Verity entity or interaction appears before testing voice, text-to-speech, or horror progression. Test one input method first. The project page says AI text can support multiple languages, while speech-to-text and text-to-speech were tested in English; if voice fails in another language, test simple English text before assuming the whole mod is broken.

After Verity appears, press Minecraft’s chat key (T by default), type Verity, say hello., and press Enter. Stable 5.7.4 handles ordinary Minecraft chat; no separate conversation command is needed. Including Verity also satisfies General > Require Verity if that option is enabled. Keep General > Immersive Mode off during diagnosis because its purpose is to hide Verity’s UI and chat.

Allow time for the first response. A local model may need to load into memory; a cloud request may be delayed by network or service limits. If the game remains responsive but Verity does not answer, check the provider connection. Once text works, hold V, speak, and release V to submit speech; M cycles available microphones. These are the defaults in 5.7.4, so check Options > Controls > Key Binds for remapped or conflicting keys.

Once typed text works, use the conversation and voice guide to test microphone permission, speech recognition, language, and audio output separately. If Verity shows a numeric provider error, use the API connection guide instead.

A short test script

Running the same few prompts each time gives you a reliable baseline after updates:

1. Verity, say hello. confirms the provider answers at all.
2. Verity, what are you doing? confirms a second request works and was not a one-off.
3. Ask a follow-up that refers to the previous answer, which shows conversational context is being kept.
4. Repeat one prompt by voice to separate speech problems from model problems.

If step 1 works but step 2 fails a few seconds later, think about rate limits or a local model still loading rather than a broken install.

Failure order that saves time

Work from the bottom of the stack upward:

1. Minecraft does not launch: verify Java runtime, game version, loader, dependencies, and duplicate files.
2. World opens but Verity is absent: verify the mod is listed as loaded, use a fresh world, check spawn or interaction instructions, and remove conflicting mods for a clean test.
3. Verity appears but has no dialogue UI: in stable 5.7.4, try ordinary Minecraft chat and turn off General > Immersive Mode. Check key conflicts before testing voice.
4. UI appears but no response arrives: verify one selected provider, API key or local service, exact model name, network access, and rate or service errors.
5. Text works but voice fails: separate microphone input, speech recognition, and text-to-speech from the language-model response. Test English text first, then one audio direction at a time.

Never publish an unredacted log until you have searched it for API keys, usernames, local paths, server addresses, and personal information. A log can be useful without becoming a secret leak.

Where to find the logs

Inside the profile’s game directory, logs/latest.log records the most recent session, and crash-reports/ contains a file for each crash. The CurseForge app and Prism expose these through the profile’s folder button. Copy only the relevant section when asking for help, and remove your API key and Windows username first.

Forge 1.20.1 versus NeoForge 1.21.1

Do not mix loaders. A Forge profile does not become NeoForge because a file is placed in its mods folder, and a release built for one loader should not be expected to initialize on the other. If you are troubleshooting 1.21.1, reproduce the issue with the recommended 1.20.1 Forge setup before concluding that the whole project is broken. That comparison separates a version-specific defect from a general setup problem.

Here, “legacy” means the older 3.4.1 NeoForge release listed on June 30, 2026. The current CurseForge description does not display the earlier “deprecated and buggy” warning. Do not treat that historical wording as a current maintainer statement or transfer Forge 5.7.x settings to this older build.

Config screen reference

Every option in Mods > Verity > Config, read from the 5.7.4 and 6.0.0-beta.9 jars, with its default. Use it to check a tutorial against your own screen.

General settings (both versions)

Option	Default	What it does
Hostile Day Threshold	5	In-game days before Verity turns into the Demon
Can Crash	On	Lets Verity kick you out of the world as a scare
True Darkness	On	The extreme darkness effect
Kill Villagers	On	Whether Verity may kill villagers
Kill Entities / Kill Wildlife	On	6.0 beta only: other mobs and random animals
Require Verity	Off	When on, a message must contain “Verity” for her to answer
Immersive Mode	Off	Hides the mod’s own interface to look like the original videos
Play Video	On	The startup video
Show Karma	On (5.7.4), Off (beta)	The karma bar above the hotbar
Custom Name, Custom Color (beta)	Verity, 0	Rename Verity or tint her (1–360 is a hue)

AI settings

5.7.4 option	Default	6.0 beta equivalent	Default
AI Provider	GROQ (or OPENROUTER)	AI Provider	OPENAI (also Groq, Gemini, Mistral, OpenRouter, Ollama)
API Key	empty	API Key	empty
AI Intelligence Level	FAST	AI Model Preset	SMART
—	—	Use Custom Model / Custom AI Model	Off
Use Ollama / LiteLLM URL / Ollama AI Model	Off / http://127.0.0.1:4000/v1/ / ollama/qwen2.5:1.5b	Provider OLLAMA + custom model	127.0.0.1:11434
Thinking Mode (Qwen3 only)	Off	Thinking Mode	On for supported models

Voice settings

5.7.4 option	Default	6.0 beta equivalent
Use text to speech	On	Use Text To Speech (TTS), On
Use Offline AI Voice	On	Text To Speech Provider: LOCAL (default), NATIVE, GROQ, KOKORO, CARTESIA
Use Native TTS	Off	(the NATIVE provider)
TTS Voice	Daniel	Voice, Daniel
Built-in speech recognition (English)	Off	Recognition Provider: NATIVE (default), GROQ, WHISPER
Use Kokoro / Kokoro Fast API URL	Off / 127.0.0.1:8880	Advanced: Text To Speech URL, Kokoro Model
Local Whisper URL	127.0.0.1:9000	Advanced: Speech Recognition URL (127.0.0.1:8080)

Two settings cause most “it worked in the video” confusion: the beta’s AI Provider default of OPENAI (set it to GROQ for a Groq key), and Require Verity, which is easy to switch on and forget: with it on, Verity ignores any message that does not contain her name.

Updating Verity safely

Updates to Verity can change configuration fields, model presets, and dependencies, as the 6.0.0 beta’s AI config rework shows. A careful update takes a few minutes:

1. Note your current working setup: Verity version, Forge build, provider, and model name.
2. Copy the whole profile, or at least back up the config folder and your worlds.
3. Replace the old Verity jar rather than adding the new one beside it.
4. Check the new release’s dependency list and update those too.
5. Launch to the title screen, reopen Mods > Verity > Config, and confirm the provider settings survived.
6. Run the short test script in a test world before returning to your main world.

Moving from 5.7.4 to a 6.0.0 beta deserves its own copied profile. Beta builds are published for testing, and their setup video differs from the stable flow.

Removing Verity JE

To uninstall, close Minecraft and delete the Verity jar from the profile’s mods folder. Remove a dependency only if no other mod in the profile needs it. Verity’s settings stay in the config folder; keep them if you might reinstall, or delete them for a completely fresh start later. If you also used Groq only for Verity, consider revoking the API key in the GroqCloud console. For local inference, stopping the Ollama and LiteLLM services frees the memory they were using.

Open a copy of a world first to confirm it loads without the mod. Worlds that never used Verity are unaffected by its removal.

Before adding Verity to a real world

Verity JE Extra is optional and separate from Verity JE. Its author requires Verity 6.0 or newer, so do not add it to a working 5.7.4 profile. Use the JE Extra guide for custom speech output and its configuration overrides. Installing an add-on is not the first repair for a failed base-mod connection.

Confirm that the game launches repeatedly, the same provider answers more than once, restart behavior is stable, and your world backup can be restored. Then add other mods gradually. Keep a record of the working Minecraft version, loader build, Verity release date, dependencies, provider, and model name. That short record is more useful than “latest everything” when a future update changes one part of the chain.