Answer ten questions about how you write and get a brand voice brief for your Instagram DM agent. The style rules arrive already counted against the 640 characters the field allows.
The two examples illustrate the voice, they are not messages to send. Every line comes from your answers through fixed rules. No AI is called.
What is a brand voice for an Instagram DM agent?
A brand voice for an Instagram DM agent is a written specification of how the agent should sound: one role, a typical reply length, a block of plain-language style rules, and real messages it can learn from. Inrō's setup gives that block a hard size: "You have 640 characters".
The same guide separates two jobs that most voice documents mix: "Style rules describe how you write. Message examples show the agent". This builder writes the rules, the part you can specify in advance. The examples stay yours, because pasting a message you never sent teaches an agent a voice you don't have.
How this builder works
How the brief is assembled
The nine-sentence style rules block
Example:
Defaults: Community Manager, normal length, warm, calm, practical, casual register, sparing emoji. The style rules read: "Write like this: warm, calm and practical. Keep it casual: contractions, short sentences, no corporate phrasing. Around 25 words per reply. Use emoji sparingly, at most one per message. Use these words: no worries, here you go, let's. Never use these words: hey guys, reach out, circle back. Answer, then invite one more question. Never promise anything you have not confirmed. Match the language the contact writes in." That is 419 of the 640 characters available.
No conversion rate appears on this page, because nobody publishes a credible conversion rate for brand voice. The numbers that do appear are anchored in Inrō's own AI agent guide or in your answers:
Watch the two word lists: they're the only input that grows without a cap of its own.
Why write the voice down
Every field an Instagram DM agent setup accepts
We read the ten most popular brand voice generators in August 2026; not one states a character limit for what it produces. Here is the field map for a real destination, an Inrō DM agent:
| Field | Where it lives | Documented limit | What it controls |
|---|---|---|---|
| Role | Voice and personality | Six presets | The job it leads with |
| Typical message length | Voice and personality | Three options, around 10, 25 or 50 words | How long each reply runs |
| Style rules | Voice and personality | 640 characters | How it writes, including emoji |
| Message examples | Voice and personality | Four to six recommended | The texture it copies |
| Account type | Goals and instructions | Seven presets | How it reads messages |
| Account description | Goals and instructions | About 120 characters | What you do, who you help |
| Goals and capabilities | Goals and instructions | Up to five, around 60 characters each | What replies steer toward |
| Audience type | Goals and instructions | About 80 characters | How much it explains |
| Other instructions | Goals and instructions | Around 640 characters | What it does or avoids |
Style rules and Other instructions share a size, not a job. Inrō's documentation draws the line: "style rules are about how the agent writes, other instructions are about what it does or avoids". This builder writes the first only.
Handling DMs by hand vs automating with Inrō
Warm, friendly and approachable all describe warmth, so the agent gets the same instruction three times.
When your style rules and pasted messages disagree, the agent leans unpredictably toward one of them.
With nothing banned, an agent falls back on service-desk phrasing and every reply reads like a form.
Choose Content Promoter and Punchy, then put your real filler words in the use list so replies sound like your captions.
Choose Lead Nurturer and Reassuring, and write the promises you cannot make into the never list before anything else.
Choose Sales Closer and Practical, then name your products in the use list so the agent never invents a product name.
Build one brief per client and keep them together, since the 640 character block is the whole voice handover per account.
A brand voice is not three adjectives. It is a block of plain instructions short enough to fit the 640 characters your agent's style rules field accepts, specific enough to argue with, and backed by examples of messages you actually sent.
Paste this into your agent
Open Inrō, go to AI Agent, then Voice and personality, and paste the block. The free plan includes 50 AI credits, 100 active contacts a month, 3 automations and 3 campaigns.
Frequently asked questions
It is a written specification of how an automated agent should sound when it answers your DMs. It has four parts in Inrō: a role, a typical reply length, a block of style rules, and message examples pasted from conversations you actually had. The first three you can decide in advance, which is what this builder produces. The fourth is evidence, and evidence cannot be generated. A brand voice document for a website argues a personality across every channel and has no size limit. A DM agent brief is narrower, because it has to fit a field and it governs one surface, the reply window.
Open Inrō, click AI Agent in the sidebar, then Voice and personality. Four things live there: the role, the typical message length, the style rules field, and the message examples list. The block this tool labels Style rules goes into the style rules field, which the documentation caps at 640 characters (updated 24 June 2026). Set the role and the length using the same names the tool gives you. The two sample replies get pasted nowhere: they exist so you can hear the voice before you commit. Inrō's documentation says changes save when you close the panel, and you can test them with the Chat with Agent button.
No. Every sentence in the brief is assembled from your ten answers by fixed rules written by hand. The same answers always produce the same brief, nothing is sent anywhere, and no model is called. That is deliberate: a tool that generates a brand voice with a language model asks a model to guess how you sound, which is the problem the brief is meant to solve. The agent you paste this into is the AI. The brief is the instruction you give it. If you would rather a model drafted one, Inrō does that inside the product: the first time you open the section it analyses your account and recent sent messages and fills in a suggestion you can edit.
From Inrō's own help centre. The article Configuring Your AI Agent's Personality and Tone, updated 24 June 2026, describes the style rules field and states the size in one sentence: "You have 640 characters." That is the whole provenance. It is not an industry standard, not a model limit, and no study says 640 characters is the right length for a voice brief. It is one product's field size, which is why it is worth publishing: it turns a vague instruction into something you can measure while you write. The same article documents a companion field, Other instructions, at around 640 characters, but that one covers what the agent does rather than how it writes.
Because three adjectives picked from one list are almost always the same adjective three times. Warm, friendly and approachable all sit on the same axis, so an agent gets one instruction wearing three hats and nothing to resolve when a reply could go two ways. This tool splits the choice across three axes: warmth, which sets how close you stand; energy, which sets how loud you are; and authority, which sets where the expertise sits. One pick per axis and the three cannot collapse into synonyms. That split is a craft judgement, not a research finding, and the notes under the formula say so.
Yes, and they matter more than the rules. Inrō's documentation puts the split in one line: "Style rules describe how you write. Message examples show the agent." It suggests four to six examples from different contexts, and the best practices article lists five worth having: a warm welcome, a pricing question, a complaint, a follow-up, and a time you had to say no. This builder cannot supply those, because they have to be messages you really sent. Copy them out of your inbox. The two examples the tool prints illustrate the voice you picked; pasting them into that list would teach the agent a voice that is not yours.
As a sentence, not a setting. Inrō has no emoji switch: the documentation says "You set emoji behaviour here, in your own words, rather than from a separate control", and gives two shapes of instruction, "Use emojis sparingly to add warmth" or "No emojis". So the emoji choice here produces no emoji characters, it produces the sentence that governs them, and that sentence is counted into your 640. Choosing Freely still bars emoji from replies about a problem, the case worth spelling out: a complaint answered with a smiley reads as a shrug. For more nuance, edit the sentence in the style rules box before you copy it.
Prose with nowhere to go. We checked ten popular brand-voice generators in August 2026: Hoppy Copy, Writecream, 1min.AI, Koala, Enji, Atom Writer, AdaptlyPost, BrandsBuilder, Backlinko's template article and Miro's workshop template. None of the ten states a character limit for what it generates. None shows a sample reply written in the voice it defines. Four mention Instagram, every one in passing. That's the gap this page fills.
No, and the difference is worth keeping straight. A script generator writes one message you send. This writes the specification an agent applies to every message it sends on its own. If you want a message for a campaign or a comment trigger, use the Instagram DM Script Generator instead. The two examples here are labelled illustrations for that reason: they show what the voice sounds like, they are not written to be sent, and they carry a placeholder price. What the agent knows is a third thing, and lives in its knowledge base, which this page does not touch.
Attract more leads, reply in seconds, and turn engagement into sales, automatically.
