SillyTavern Author's Note Explained & a Simpler Alternative

SillyTavern's Author's Note is a dedicated text field accessible from the "Advanced Formatting" or "Author's Note" panel in the SillyTavern UI. When you type something into this field, it gets injected into the prompt sent to the language model — specifically, it is placed at a configurable position relative to the character card, scenario, and chat history. By default, the note appears at the very top of the prompt, before any other context. This prime position means the model treats the note as a high-priority instruction. Unlike a character's description or example messages, the Author's Note is global across all chats unless you set it per-character. It persists across sessions until you change it. The note supports a character limit (typically 500 tokens or less, depending on your model's context window) and can include formatting like bold, italics, or even JSON-style instructions. Because SillyTavern sends the entire context with every request, the note is applied to every message you send, making it a persistent steering mechanism.

SillyTavern provides multiple ways to influence model behavior: the character card's description, the system prompt, and the Author's Note. The character description is tied to a specific persona and defines that character's core traits, background, and voice. The system prompt (set in the "System Prompt" field) is a global instruction that applies to all characters and is usually placed near the top of the context. The Author's Note, however, is unique because it is injected at a position you choose (top, bottom, or relative to other elements) and is meant for temporary or session-specific adjustments. For example, if you want a character to adopt a more formal tone for a single scene without editing their permanent card, the Author's Note is the right tool. It also supports "injection frequency" — you can set it to appear every N messages rather than every message, which reduces token usage and prevents oversteering. In contrast, the system prompt is always sent. The Author's Note is also a community favorite for adding subtle atmospheric cues or genre-specific rules (e.g., "This is a noir detective story. Use period-appropriate slang.") without cluttering the character card.

SillyTavern gives you two critical controls for the Author's Note: depth and position. Depth controls where the note sits in the prompt order relative to other context chunks. The default depth is 0, meaning the note is placed at the very top of the prompt (highest priority). Higher depth values (e.g., 4) push it further down, placing it after the character card, scenario, and maybe a few chat turns. This is useful if you want the note to act as a subtle reminder rather than a commanding instruction. Position options include "before character card", "after character card", or "at the end of the context". Most users keep it at the top for maximum impact, but for complex roleplays, placing it after the character card avoids overriding the character's core identity. The injection frequency setting (default: 1) determines every Nth message the note is included. Setting it to 2 means the note appears every other message, saving tokens and reducing repetition. For long-running chats, frequency 3-5 is common. You can also set the note to only inject when the context length exceeds a certain threshold — a smart way to maintain coherence without wasting tokens on short exchanges.

An effective Author's Note is concise, specific, and action-oriented. Avoid vague instructions like "be more interesting"; instead, write something like "The character speaks in short, clipped sentences. They avoid eye contact and fidget with their sleeves." Using formatting like asterisks for actions or quotes for dialogue helps the model parse intent. Many experienced users employ a structured format: start with a one-line directive, follow with 2-3 behavioral bullet points, and optionally include a sample response. For example: "*Scene: Tense interrogation*\n- The detective remains calm but suspicious.\n- Every answer leads to another question.\n- 'I've seen your kind before,' he says, leaning forward." Keep the note under 300 tokens to avoid eating context. Avoid contradictory instructions: if the character card says "friendly and open", don't write "be cold and distant" in the note — the model may average the two. Test different phrasings; the same instruction phrased as a command ("Speak formally") vs. a description ("The character uses formal language") can yield different results. Also, remember that the note affects every message, so avoid anything that would break immersion if repeated.

The Author's Note shines when you need to temporarily alter a character or scene without permanent edits. A frequent use case is genre shifting: if your fantasy roleplay suddenly enters a horror segment, you can set the note to "The atmosphere becomes eerie. Shadows lengthen. Unseen things whisper." This instantly recontextualizes the model's output. Mood control is another: if a character is supposed to be drunk, angry, or grieving, the note can enforce that emotional state across responses. Temporary traits — like a character who is cursed, injured, or under a spell — are also handled cleanly. The note is also used for meta-instructions like "Do not narrate the user's actions or thoughts" or "Keep responses under 100 words." Some users employ it to simulate a narrator or third-person voice. In group chats, the note can set the overall scene or enforce group dynamics (e.g., "The party is in a crowded tavern. Characters speak over each other occasionally."). Because the note can be toggled on/off per message, it's easy to experiment without commitment.

SillyTavern's Author's Note is powerful but has drawbacks. It requires manual configuration, consumes context tokens, and can become repetitive if not tuned. For users who prefer a simpler experience without managing API keys, context limits, or injection frequencies, AIAngels offers a built-in memory system that achieves similar persistent steering without any setup. AIAngels' permanent memory remembers every detail from conversation 1 and applies it naturally — no need to write a note. Its custom companion builder lets you define appearance, personality, and interests upfront, and the model maintains those traits without extra prompting. For users tired of tweaking depth settings and token budgets, AIAngels provides unlimited text chat on the free tier and premium plans starting at $2.99/mo. While SillyTavern is great for power users who want fine-grained control, AIAngels removes the overhead entirely, making it ideal for those who just want consistent, immersive conversations without the learning curve.