Skip to main content

🎙️ Brand Voice

Brand Voice shapes how drafts are written for your X account — suggested replies, Slack drafts, and generated posts all follow these settings. Set the tone once, add your facts and guardrails, and every AI-written draft sounds like you.

Overview​

  • Pick a tone and reply length that fit your account
  • Give the writer facts it can rely on and voice samples to imitate
  • Set hard guardrails: things to never say, plus emoji, hashtag, and language policies
  • Optionally ground drafts in your knowledge base so replies answer with real information
  • Test drive the voice against sample mentions before it writes anything real

Configuration​

To configure Brand Voice:

  1. Open your project in the Fren One webapp.
  2. Select the X account you want to configure.
  3. In the sidebar under Agent Configuration, select Messages, then Brand Voice.

Brand Voice is visible to the project Owner and Project Admins only — other members don't see the page.

Voice & persona​

These settings define how drafts sound.

OptionDescription
ToneCasual, Professional, Technical, Friendly, Empathetic, or Authoritative.
Reply lengthConcise, Balanced, or Detailed. "Concise suits X best."
About this accountFacts the writer can rely on — what you do, who you serve, what you offer. Max 2,000 characters.
Voice samples"Paste 2–3 of your best posts. Drafts imitate their rhythm and vocabulary." Max 2,000 characters.

Guardrails​

Guardrails are hard limits the writer must respect, regardless of tone.

OptionDescription
Never sayOne rule per line — topics, claims, or phrases drafts must avoid. Overrides everything. Max 2,000 characters.
EmojiNever, Sparingly, or Freely.
HashtagsNever or When relevant.
LanguageMatch the post or Always English.

Knowledge base grounding​

Toggle "Ground drafts in your knowledge base" to pull facts from your project's knowledge base so replies answer with real information. With grounding on, each draft looks up relevant knowledge-base entries before writing — and each draft consumes AI credits.

Writing model​

Choose which model writes your drafts:

OptionDescription
FastLowest credit use.
Quality"Best everyday writing. The default."
Reasoning"Careful replies for tricky threads."

Test drive​

Try the voice before it writes anything real:

  1. Pick a mode: "Reply to a mention" or "New post".
  2. Choose a scenario pill — Recent mention, Support question, Praise, or Complaint — or paste your own sample into the textarea (max 1,000 characters).
  3. Select "Generate sample".
Test drives are metered

Each sample is a real draft generated with your current settings — it costs AI credits just like a production draft.

How drafts are written​

Every draft produced through Brand Voice follows the same runtime guardrails:

  • Drafts are plain text, capped at 280 characters.
  • The draft follows your emoji, hashtag, and language policies, and stays within your chosen reply length.
  • The writer is instructed to never invent facts, prices, dates, or promises.
  • "Never say" rules are enforced last, with a final re-check before the draft is returned.

These settings power the drafts you see in inbox suggested replies, the Slack "Generate reply" action (see Integrations), and the test drive above. AI Workspace chat drafts are a separate path that does not apply Brand Voice — see AI Post Drafting.

Saving your changes​

This page uses a batched save — edits stay local until you select Save in the bar that appears at the bottom of the page. Leaving the page with unsaved changes shows a warning first.

Best practices​

A voice that holds up
  • Fill in About this account with concrete facts — it's the difference between a generic reply and one that answers correctly.
  • Paste real posts into Voice samples; drafts imitate their rhythm and vocabulary, so pick your best.
  • Keep Never say rules short and unambiguous — one clear rule per line beats a paragraph.
  • Turn on knowledge-base grounding if members ask factual questions, and keep your knowledge base current.
  • Use the test drive with the Complaint scenario — the hardest case is the best check of your guardrails.
Troubleshooting

Drafts aren't using your knowledge-base facts?

  • Check that "Ground drafts in your knowledge base" is toggled on and the knowledge base actually contains the facts.
  • If a knowledge-base lookup fails, the draft falls back to writing without it rather than failing entirely.

AI Workspace drafts ignore these settings?

  • That's expected — workspace chat drafts are grounded in community data but don't apply Brand Voice. Edit them in the Tweet Scheduler, or draft from a Brand Voice surface instead.

You can't see the Brand Voice page?

  • Brand Voice is limited to the project Owner and Project Admins.

On this page