Settings¶
Open LIMA settings from the File menu, then Settings.
Appearance¶
| Setting | Description |
|---|---|
| Text Size | Font size for the chat history. Options: 10–30 pt. Default: 14 pt. |
Audio¶
| Setting | Description |
|---|---|
| Language | Sets the language for both voice input and AI responses. Options: English, Vietnamese, Tagalog. |
| Text-to-Speech voice | The voice LIMA uses for its responses. See the voice list below. |
| Speech rate | How fast LIMA speaks. Options: slow, medium, fast, fastest. |
| Text-to-Speech mode | How LIMA generates its voice. Online (Conversational voice) uses a natural-sounding voice and starts speaking almost immediately — higher quality, but requires an internet connection. Offline (Robot Style voice) uses a voice built into your computer — it works without internet and keeps speech on your device, but sounds more robotic. |
| Microphone sensitivity | How sensitive the microphone is to your voice. Increase if LIMA misses commands; decrease if it triggers accidentally. |
| Finished-speaking sensitivity | How quickly LIMA decides you have finished talking and starts to reply. A lower setting makes LIMA respond sooner but it may cut you off; a higher setting makes it wait a little longer to be sure you're done. Default: 0.75. If LIMA often replies before you have finished, for example because you pause to think, speak with an accent, or speak in a second language, raise this setting a step at a time. |
Available voices¶
| Voice | Description |
|---|---|
| Despina | Chill Young Female |
| Enceladus | Middle-Aged Male |
| Rasalgethi | Chill Young Male |
| Laomedeia | Professional Woman |
| Algenib | Australian Man |
| Aoede | Australian Woman |
| Sadaltager | British Man |
| Callirrhoe | British Woman |
Note
Algenib, Aoede, Sadaltager, and Callirrhoe are available for English and Vietnamese only. Tagalog uses Despina, Enceladus, Rasalgethi, and Laomedeia.
Hot keys¶
Six hotkeys can be customised here. See Keyboard Shortcuts for the full list and instructions.
Features¶
| Setting | Description |
|---|---|
| Descriptive Video Service (DVS) | Automatically narrates what is happening on screen during YouTube and Facebook videos. Options: off, minimal, regular, verbose. Default: regular. |
| Dynamic Website Narration (DWN) | Automatically reads out changes on websites as you browse. Options: off, minimal, regular, verbose. Default: regular. |
| Screen to capture | Which display LIMA screenshots when observing your screen. Options: Main screen only, All screens (span every monitor). Default: Main screen only. |
| AI model | The AI that powers all of LIMA's responses and screen reading. See the model list below. |
| Thinking level | How much the AI model reasons before answering. Off is fastest; higher levels reason more deeply but respond more slowly. Options: off, low, medium, high. Default: low. |
| Conversation history (messages) | How many of your recent messages LIMA keeps in mind during a conversation. Options: 10, 15, 20, 30, 50. Default: 20. |
| Screenshot count | How many recent pictures of your screen LIMA looks at when answering. More can improve accuracy but is a little slower. Options: 1–5. Default: 2. |
| Maximum agent steps | How many steps LIMA may take to finish one request before it stops and tells you. Raising it lets longer tasks finish, but setting it too high risks getting stuck in a loop. Options: 10, 20, 40, 80, 120, 200. Default: 80. |
| Enable voice mode on startup | When enabled, voice mode activates automatically when LIMA starts. |
| Minimize to tray on close | When enabled, closing the LIMA window sends it to the system tray instead of exiting. |
Available AI models¶
| Model | Plan required | Notes |
|---|---|---|
| Qwen3.8 (Qwen) | All plans | Fast and capable. The default model on every plan. |
| Opus 5.5 (Anthropic) | Pro | Complex tasks; consumes usage quickly. |
| GPT-5.5 (OpenAI) | Pro | Most capable, slower. |
| GPT-5.6 Luna (OpenAI) | Pro | Fast and capable. |
| Kimi K2.6 (Moonshot) | Standard or higher | Fast and smart; thinks before answering. |
| V4.1 Flash (DeepSeek) | Go or higher | Cheap and fast, good at everyday tasks. |
| Muse Glimmer (Meta) | All plans | Small and fast, good at everyday tasks. |
Existing Opus 4.8 selections automatically use Opus 5.5.
Memories¶
Memories are not set in Settings. LIMA holds up to 10 memories; adding one when the list is full replaces the oldest. See Memory.