Chat overview
A map of the chat interface — every control, what it does, and the docs page that explains it in depth.
The chat screen is intentionally minimal. Most of the time, the only thing on it that matters is the input box. But behind a few unobtrusive icons sits the full feature set — forking, group chats, voice, plugins, model switching, identities, memory.
This page is the map. Each section links to a deeper guide.
The chat layout
The strip of icons on the right opens the side panel — six tabs that hold the character profile, gallery, scenarios, memory, plugins, and per-chat settings. Side panel & chat settings →
Every control, mapped
| Control | Where | Does what | Deep dive |
|---|---|---|---|
| Model picker | Top of chat | Switch LLMs mid-thread. Multiplier shown next to each. | Choosing a model |
| Reasoning toggle | Inside the model picker | Show or hide AI thinking process. Per model, per chat. | Reasoning mode |
| Identity selector | Top right | Which persona is speaking. Each remembers separate memories. | User identities |
| Voice call | Phone icon, top right | Switch from typing to real-time voice. | Voice calls |
| Plugins (🧩) | Above the input | Inject behaviour, run tools, or add interactive buttons. | Plugins |
| Scenarios | Side panel | Pre-built contexts the creator authored (e.g. "Coworker", "Rival"). | Scenarios & memory |
| Memory | Side panel | View, edit, or pin what the character remembers about you. | Memory |
| Smart suggestions | Above the input | Scenario starters when empty, completions while typing. | Smart suggestions |
| Generate first message (✨) | Input row, new chats only | Author or AI-generate the opening message / scene before you start. | Below |
| Voice input (🎙) | Input row | Push-to-talk speech-to-text. Long-press opens voice settings. | Voice calls |
| Side panel | Tabs, top right | Info, gallery, scenarios, memory, plugins, and per-chat settings. | Side panel |
| Fork | Long-press any message | Branch the conversation. Both halves survive. | Forking |
| Regenerate (↺) | Below AI replies | Roll another version. Long-press to retry with a different model for this one reply. | This page |
| Edit | Long-press your messages | Change what you sent. Reply rebuilds from there. | This page |
| First-response enhancement | After the first AI reply | One-tap rewrites for more dialogue, action, intimacy, etc. | First response |
| Dual-model first-reply comparison | First reply of a new chat | Two models reply side-by-side; pick the winner, feed community stats. | Dual comparison |
| Peek into mind | Plugins toolbar | Generate diary, dreams, unsent letters, inner thoughts. | Peek into mind |
Hidden long-press actions. Two controls do extra duty when you press and hold instead of tapping: Regenerate opens a model picker so you can retry that single reply with a different LLM, and the voice (🎙) button opens voice settings. Long-press also opens Fork on any message and Edit on your own.
The three things you'll do in every chat
1. Send a message
Type and hit return. Empty input shows scenario suggestions; typing shows guided completions. Either is optional.
2. Steer when something feels off
You have four moves before you give up on a reply:
- Regenerate — same input, different roll.
- Edit your message — change the prompt, then regenerate.
- Switch model — the model picker is two clicks away.
- Fork — keep the bad reply for posterity, branch off and try something different.
3. Save what mattered
Open the memory panel in the side panel and add an entry for anything you want preserved. Manually-added memories survive every summarization and the character references them forever. See Memory.
Generate the first message
On a brand-new chat, a ✨ button sits in the input row. It opens the first-message generator, which sets the opening scene before you type anything.
- Enter a prompt to steer it ("A fantasy adventure where we meet in a tavern...") — or leave it blank for a random scenario.
- The AI writes the character's opening message from your prompt, plus the active scenario, moment, and identity. Preview it, edit inline, then Apply.
- Prefer to write it yourself? Flip "Use my input directly without AI optimization" and your text becomes the opening message verbatim.
- In voice mode, the generated opener is dialogue-only so it reads aloud cleanly.
This is different from first-response enhancement, which tweaks the AI's reply after you've sent your first message.
The side panel
The tab strip on the right opens a six-tab panel: info, gallery, scenarios, memory, plugins, and a settings tab with per-chat Advanced Settings (model, voice, display, data). Full breakdown: Side panel & chat settings.
Group chats
The same screen, but with more than one character at the top. Reverie's orchestrator decides who speaks next based on conversation flow. You can also @-mention a character to force them to respond.
Two stories worth knowing
Long conversations stay coherent. Reverie has a three-layer memory system: live context, rolling summary, and pinned long-term memory. Threads with thousands of messages still remember your first kiss and your dog's name. Memory →
Forking is the safety net. You never lose a conversation by trying something risky. Fork before you say the bold thing — both branches will be there tomorrow. Forking →