Skip to main content

Brand Voice & AI Writing

Joryio's AI writes marketing copy that sounds like your brand - not like a generic assistant. The Brand Voice profile is how you teach it.

Find it under Brand Studio → Brand Voice (at /brand/voice).

Think of it as two halves of one identity: the voice is how your brand sounds (tone, vocabulary, examples), and the Brand Kit inside it is how your brand looks (logo, colors, font). Every AI writing surface in Joryio reads the voice; every AI design surface reads the kit.

Quick start​

Five minutes of setup pays off across every AI feature:

  1. Open Brand Studio → Brand voice and click New voice. Pick a starting point: From your website (recommended - one read fills the brand kit and lets the AI learn the voice from your own pages), From texts you have written, or Describe the brand.
  2. Review the draft in the editor. Every section the AI filled is marked until you have looked at it; nudge the sliders, prune the lists, fix anything it got wrong.
  3. Open Test the voice, write a one-line brief and press Write it. The draft is checked against the voice's own words and rules. Refine the profile, not the prompt.
  4. Press Create voice. Nothing is saved before that.

Multiple voices​

A workspace can hold several voices (agencies, multi-brand companies) - one is the default and every AI-written text follows it unless you pick another. The assistant knows all your voices: name one ("write an email for Datavera") and it writes in that voice and says so; if it's ambiguous it asks once. The AI-suggest panels in the editors show a voice picker only when more than one voice exists.

What a brand voice contains​

  • Tone sliders - formality (casual ↔ formal), playfulness (serious ↔ playful), and urgency (relaxed ↔ urgent), each 1–10.
  • Tone characteristics - short adjectives the writing must embody ("warm", "direct", "premium").
  • Words to use - signature words and phrases woven in naturally.
  • Words to avoid - strictly forbidden. The AI is instructed never to use them, and Joryio double-checks every output on the server; anything that slips through is flagged before you see it.
  • Reference examples - up to 3 texts the output should sound like (the "same author" test).
  • Guidelines - free-form do's and don'ts.
  • Language settings - the voice's primary writing language, which drives the default generation language when a request doesn't specify one.
  • Hebrew settings - audience form (masculine / feminine / mixed-neutral), register (formal / natural / everyday), and niqqud. Grammatical gender matters in Hebrew copy, so the voice states it explicitly instead of leaving the model to guess. Hebrew copy is generated with the model that performs best for Hebrew.

Each workspace has one default voice that all AI writing uses automatically; you can create several and pick per generation.

Tell the AI what your company is​

Each voice carries a business category (e-commerce, gaming, insurance…) and a short "what does the company do" description. These are injected into every AI generation and every Assistant conversation, so suggestions are framed for your kind of business - cart and shipping language for a store, trust and clarity for an insurer. The auto-analysis below fills them for you; you can always adjust.

Creating a voice​

New voice is the one entry point. It offers three starting points, and each ends in the same editor with a draft you review before saving:

  • From your website (recommended) - Joryio reads your home page and a few of its pages. One read fills the brand kit (logo candidates, colours, font, social links) and gives the AI real copy to learn the voice from. Tick Also learn from texts I have written to add pasted texts on top - useful when your emails sound different from your site. If a site blocks automated visitors, Joryio says so and you can paste texts instead. The message names the reason: the site blocks automated reading (many large retail sites answer 403 or 429 to anything that is not a browser - that is the site's choice, not a fault in Joryio), it did not answer in time, or it answered an error such as 404.
  • From texts you have written - paste 2-10 of your best texts (emails, posts, SMS), one per blank line. The AI drafts the full profile: slider positions, personality, words to use and avoid, and sample texts pulled from your own copy.
  • Describe the brand - no copy yet? Answer four questions (the first two are required) and the AI designs the voice, including two sample texts written in it so you can judge it immediately.

Start blank instead opens an empty editor. Either way the result is a draft: nothing is saved until you press Create voice.

From a document​

Paste a document is the fourth starting point, for writers who already have their voice written down: one document with example posts or messages you wrote, your signature phrases, and a profile of how you write (who you are as a writer, how you open, what you never do). Paste it, or upload a .txt or .docx file (up to 5 MB; for a PDF, copy its text and paste it). Up to 100,000 characters are read.

One analysis splits the document into its parts:

  • Your posts - up to 20, kept word for word. The AI never re-types them: it points at where each post starts and ends and Joryio copies your exact text, in its original language. When it writes for you it picks the two or three that fit the message (short ones for SMS, longer ones for email) as style references, and never copies their content.
  • Signature phrases - your own expressions, kept only when they really appear in the document. The AI uses one or two where they fit naturally and never forces them in.
  • The profile - your writer profile becomes the voice's tone sliders, personality, words to avoid and rules.

You review all of it in the dialog and can edit or remove anything. If the document leaves something out that would change the writing - words or claims you never use, how SMS differs from email, who you write to, which language or dialect - the AI asks up to three follow-up questions, chat style. Answer what you can and Skip the rest, then press Update draft. The draft then opens in the editor, and nothing is saved until you press Create voice.

From the Assistant. You can do the same in the AI Assistant chat: paste the document (or attach a .txt or .docx file) and ask for a brand voice. It runs the same analysis, tells you what it kept, asks the same follow-up questions in one message, and shows the result as an Apply card. Pressing Apply saves the voice; Open in the editor opens the same draft in the Brand voice editor for detailed editing.

The editor​

A voice is edited on a full page with a rail of sections: Basics (name, category, writing language, description, default), Tone (three sliders, each with an example sentence that changes as you move it, plus personality traits), Vocabulary (words to use, words to avoid), Examples & rules (up to three sample texts and the rules every message follows, as separate rows), Brand kit (the look, with a live email preview) and Test the voice.

Test the voice drafts a promo email, an SMS or a push from a short brief, in the voice as currently edited - saved or not - and checks it against that voice's own lists: avoided words that slipped in, brand words used, rules written to. Nothing is sent or saved.

The default voice cannot be deleted while another voice exists; set another voice as default first. Duplicate copies a voice, brand kit included, as a starting point for a second brand.

The voice keeps learning​

Joryio quietly improves each voice over time. When you accept, edit, or ask the AI to refine generated copy, those signals are periodically distilled into short, durable learnings - rules like "never opens with a question" or "prices always include VAT" - that are re-injected into every future generation for that voice.

Learnings are visible on the voice profile, and each one can be edited or deleted if the AI drew the wrong conclusion. They accumulate slowly and deliberately; a one-off edit won't rewrite your voice.

Brand Kit - your visual identity​

The Brand Kit card inside each voice holds how your brand looks: logo, colors, and font. The AI uses it whenever it builds an email template for you - logo in the header, buttons and accents in your colors.

A kit contains:

  • Logo URL - used in generated email headers.

  • Colors - primary (buttons, links, headline accents), secondary accent (badges, dividers), email background, body text, and the text color used on primary-colored buttons.

  • Palette swatches - extra colors found during website extraction, shown as clickable chips; click one to promote it to your primary color.

  • Font - the preferred font stack for generated emails (with a safe system-font fallback).

  • Header & footer HTML (optional) - verbatim header/footer markup lifted from your real emails. When present, generated templates sandwich their content between these for the highest-fidelity branding.

  • Social links (optional) - real profile URLs for Facebook, Instagram, LinkedIn, X, TikTok and YouTube. Auto-fill finds them on your site; generated footers carry only the ones that are set, as icons with real links. The company address that footers also carry lives in Settings → General, not in the kit.

Filling the kit​

  • From your website - when a voice is created from a website, the kit is filled at the same time. Later, the Re-read website button on the Brand kit section refreshes the logo, colours, font and social links; your voice settings are never touched. Extracted colours appear as swatches; click one to make it your primary colour. If a site blocks automated visitors, Joryio degrades gracefully: it can usually still offer a logo candidate, and you set the colours and font manually.
  • Manual - upload a logo or pick one from your media library, or paste a logo URL, and pick the colours with the colour pickers. A logo taken from your site or pasted as a URL is copied into the media library when you save, so it stays available if the site changes. The email preview beside the kit shows a whole header, headline, button, link and footer in your kit, so colours are judged together.

Which voice a generation uses​

When several voices exist, the pick order is simple and predictable:

  1. An explicit pick wins. The AI-suggest panels show a voice picker whenever more than one voice exists; whatever you select there is used for that generation.
  2. Naming a voice in chat counts as a pick. Tell the Assistant "write it for Datavera" and it uses that voice (and says so). If the request could match more than one voice, it asks once instead of guessing.
  3. Otherwise, the workspace default applies. Mark the right voice as default and single-brand teams never have to think about it.

Where the voice (and kit) apply​

One profile, many surfaces - configure it once and every AI feature stays on brand:

  • AI subject lines in the campaign editor - see below.
  • AI-written message copy - email bodies, SMS, and refinements requested from the Assistant all follow the voice's tone, vocabulary, and forbidden-words list, and respect channel length limits.
  • Assistant-proposed email templates - ask the Assistant for a template and it writes the copy in your voice, then assembles a branded, email-safe design from your Brand Kit. Applied templates are saved as regular templates you can fine-tune.
  • Assistant-drafted journeys - when the AI Assistant drafts a journey for you, every email step inside it is generated in your voice and pre-wrapped in your Brand Kit's header, colors, font, and footer - the draft opens in the drag & drop editor already populated and on brand.
  • AI in-app messages - Assistant-generated in-app templates use the kit's logo and colors for a branded modal out of the box.

AI email templates​

Ask the Assistant for a template - for example, "create an email template announcing our summer sale, 20% off with code SUN20". It shows an Apply card; on Apply, the AI writes the copy in your brand voice and Joryio assembles a clean, email-safe design from your Brand Kit. The result is saved as a regular template you can open and fine-tune in the template editor. Nothing is created before you press Apply.

The division of labor is deliberate: the AI writes copy only, and deterministic template code assembles the HTML from your kit's tokens. That's why AI-generated emails never come out with broken layouts or off-brand colors - the design isn't hallucinated, it's assembled.

Branded sections & product blocks in the editor​

The email editor has a Branded sections block category - hero, paragraph, button, coupon code, divider, and products - every block already rendered in your Brand Kit colors and font when you drag it in. The product block has a Choose product action that opens your synced store catalog (Shopify/WooCommerce): pick a product visually and its real image, name, price, and link fill in - no code or variables needed. The AI uses these exact same sections when it generates templates, so AI-built and hand-built emails always match. (In SMS, where there are no blocks, product references still use Liquid variables.)

AI subject lines in the campaign editor​

When editing an email variant, click AI suggest next to the subject field:

  1. Pick the language and (optionally) a brand voice - the workspace default is used otherwise.
  2. Describe what the email is about in one sentence.
  3. Get several genuinely different angles (benefit, curiosity, urgency…), each with a matching preheader. Click one to apply it.

Subject lines respect proven limits (≤60 characters recommended) and your forbidden-words list. Variants pair naturally with A/B testing - generate a few angles and let the experiment decide.

FAQ​

Does any of this send messages by itself? No. The voice and kit shape what the AI writes and designs; everything it produces is a draft, a template, or a suggestion you apply. Sending stays in campaigns and journeys, under your control.

We have no website - can we still use the Brand Kit? Yes. Auto-fill is a convenience, not a requirement - set the logo URL and pick the colors manually, and the live mini-preview shows you the result immediately.

Can different journeys or campaigns use different voices? Yes. Any AI generation can pick a non-default voice, and copy generated for one journey doesn't bind the journey to that voice - the pick happens per generation.

Someone edited the voice - does existing copy change? No. The voice is read at generation time. Copy already written and saved stays as-is; the next generation follows the updated profile.

Privacy​

Your prompts and sample texts are used only to generate your content. Generation metadata (token counts, cost, whether a draft was accepted) is stored for quality and billing; the generated text belongs to you.