Brand Voice

Brand voice tells Amplicast's AI how to sound for a specific project. It's the difference between generic "ChatGPT-flavored" content and content that actually sounds like your brand.

Voice is per project, not per user — so TIRIDA can sound playful while your B2B SaaS sounds clinical, even though both run in the same dashboard.

What's in brand voice

Field	What it does
Publication	The brand name as it appears in voice ("Acme Daily", "@yourname")
Description	1–2 sentences: what the brand is
Tone	How content should sound ("Playful, warm, storytelling-first")
Perspective	POV ("First-person creative journal" / "Third-person editorial")
Audience	Who's reading ("3D artists and character designers")
Avoid	What NOT to do ("No corporate jargon, no AI-hype-speak")
Signature phrases	Phrases to weave in naturally
Default word count	Target length per piece
Default image style	Visual aesthetic for AI-generated images
Default target platforms	Where this content typically posts

Amplicast prepends these fields to every AI prompt for that project — both for autopilot generation and for the source-to-platform transform pipeline.

Three ways to set voice

You pick whichever fits your workflow. They all populate the same fields and you can mix freely.

1. Sync from a brand hub (fastest if you have one)

If your brand identity already lives in an external system that exposes a public API matching our brand hub schema, Amplicast can pull it in one click:

Open Project Settings → Voice
Click Sync from brand hub
The voice form fills in automatically

Override the lookup slug. If your project's slug here (e.g. my-brand) differs from the slug on the brand hub (e.g. mybrand), set the override in General → Brand hub slug.

Pointing at your own brand hub instead of the default: this is a deployment config — BRAND_HUB_BASE_URL. Self-hosted Amplicast clients can change it; on the SaaS plan, reach out at hi@amplicast.io and we'll connect your hub.

2. Paste a brief, let AI extract (fastest if you write your brand briefs in prose)

The friendliest path for most users:

Open Project Settings → Voice
Expand "Or paste a brief and let AI extract the fields"
Paste 1–2 paragraphs describing your brand — what it is, the tone, who it's for, what to avoid, signature phrases, anything else relevant
Click Extract
The 9 fields populate. Each field the AI confidently extracted gets a violet "AI-extracted" badge so you can see at a glance what to review.
Edit anything that needs tweaking. Click Save voice.

The AI deliberately leaves a field empty rather than guess. So if your brief didn't mention default word count, that field stays blank for you to fill manually (or skip — empty fields are just not added to the prompt).

3. Fill the form by hand (when you want full control)

Open Project Settings → Voice and type into each field. Save when done.

You can also use this path after syncing from the hub or AI-extracting from a brief — tweak the fields that didn't come out right, then save.

How voice shows up in generated content

The voice block is prepended to every AI prompt for the project. It looks roughly like this:

BRAND VOICE FOR THIS PROJECT — match this consistently:
Publication: <your publication>
Description: <your description>
Tone: <your tone>
Audience: <your audience>
AVOID: <your avoid list>
Signature phrases (use naturally, don't force): <your phrases>
Target length: ~<your word count> words

Then the actual content generation instructions follow. So the AI hears the brand identity first, then the task ("write a LinkedIn post about X").

Brand voice vs. Writing Style vs. Reference Library

Three things that look similar but serve different roles:

	Per	Holds	Status
Brand voice	Project	Tone, audience, avoid, signature phrases, image style	New (this page)
Writing Style	User	Concrete past writing samples, words to avoid, default image prompt	Existing — complements voice
Reference Library	User	Uploaded logos and reference images	Existing — visual inputs for image generation

When the AI generates content:

Brand voice comes first — it sets the brand identity
Writing Style writing samples come next — concrete examples of how you write
Reference Library images feed into image generation for visual consistency

If a field overlaps (e.g. brand voice's "avoid" and writing style's "words to avoid"), brand voice wins. Writing samples are always included because they complement voice with examples, not duplicates.

Tip: leave the per-user Writing Style to your own examples of good writing, and use Brand Voice for the brand-level constraints. They layer cleanly.

Reference images in scheduled publishing

By default, your Reference Library (uploaded reference images + logos in Writing Style settings) influences AI image generation for Autopilot and Image Preview only. Scheduled publishing — the path that runs when a source-fetched item or approved review draft gets transformed for each platform — uses the AI image generator's text-prompt-only mode.

You can opt in per project to extend reference-image influence to scheduled publishing too:

Open Project Settings → Logo and brand assets
Tick "Apply reference images to publishing"
Save

When the toggle is on AND your Writing Style has at least one reference image uploaded, scheduled posts in the ai-generate image mode go through the same Gemini Vision reference analysis that already drives Autopilot. You'll see 🎨 Using N reference images from library in the worker logs for each post processed.

Important caveats:

Off by default so existing scheduled-publishing behavior is unchanged for projects that have never set the toggle.
The setting uses the reference library state at the moment each post is transformed — re-transforming an old draft after editing your library applies the new state.
Only AI image generation is affected. Non-AI image transforms (resize, filter, watermark, logo overlay) are untouched.
Toggling it off doesn't retroactively undo any image generated while it was on.
Per-content-type control (e.g. "use refs for square images but not portraits") isn't in v1 — it's a single per-project switch.

When to use which path

Situation	Recommended path
Your brand already has a published API / brand hub	Sync from hub
You write your brand briefs as prose (e.g. in a Notion doc)	Paste + AI extract
You're refining specific fields, or starting from scratch with full control	Fill the form by hand
You synced from the hub but want one field tweaked	Sync, then edit that field, then Save

FAQ

My brand voice is mostly visual — colors, logos, image style. Does that work? Yes. The Default image style field gets fed to image generation. The Publication + Description + Tone fields keep the text on-brand too. Logos go in Project Settings → Logo (separate from voice), and reference images go in Reference Library.

Can I have different voices for different content types (e.g. quizzes vs. stories)? Today voice is one-per-project. If you publish multiple content types with different voices, the brand hub schema supports image_style_by_type for visuals; text voice is uniform. If you have a real use case for per-content-type text voice, let us know at hi@amplicast.io — it's on the roadmap if there's demand.

My voice doesn't seem to be applied to a generated post. Two things to check:

The autopilot topic is scoped to the right project — old topics created before the project layer have no project_id and skip voice. Recreate or edit them via the autopilot tab.
The voice has at least one field filled. Empty voice is a no-op.

Can voice be language-specific (EN + DE)? Today it's one block per project. For bilingual brands, the cleanest pattern is two projects (mybrand-en, mybrand-de) with separate voice on each. The brand hub schema already supports locales — full bilingual voice within a single project is a future improvement.

Technical: brand hub schema

If you're integrating your own brand hub for SaaS clients (or for your own multi-deployment setup), the API contract Amplicast expects:

Endpoint: GET <BRAND_HUB_BASE_URL>/api/public/brand?slug=<brand-slug>&include=voice

Response shape:

{
  "ok": true,
  "brand": {
    "slug": "your-brand-slug",
    "name": "Your Brand",
    "voice": {
      "publication": "...",
      "description": "...",
      "tone": "...",
      "perspective": "...",
      "audience": "...",
      "avoid": "...",
      "signature_phrases": ["..."],
      "default_word_count": 0,
      "image_style_default": "...",
      "image_style_by_type": { "type-key": "style description" },
      "default_target_platforms": ["instagram", "tiktok"],
      "enabled_content_types": ["..."],
      "default_content_type": "...",
      "seo_title_template": "{{title}} — Brand",
      "seo_keywords_default": ["..."]
    }
  }
}

All voice fields optional. Auth is configurable (none / API key); CORS not required (Amplicast calls server-to-server).

For integration help or setting BRAND_HUB_BASE_URL on the SaaS plan, contact hi@amplicast.io.

What's in brand voice​

Three ways to set voice​

1. Sync from a brand hub (fastest if you have one)​

2. Paste a brief, let AI extract (fastest if you write your brand briefs in prose)​

3. Fill the form by hand (when you want full control)​

How voice shows up in generated content​

Brand voice vs. Writing Style vs. Reference Library​

Reference images in scheduled publishing​

When to use which path​

FAQ​

Technical: brand hub schema​