Give your tutor a voice, a face, and a manner

2026-07-07 · The Alltutors.ai team

What you'll build

This guide gives your tutor a face, a voice, and a manner that stay consistent across every lesson. Think of it as an AI version of you. It looks and sounds like how you teach, and it works as an extension of you, never a replacement. None of this is cosmetic. The persona settings you pick change how the tutor explains things and how it reacts when a learner gets something wrong. That's the part learners feel first.

Step 1: Know which settings change behavior

The voice and avatar decide how the tutor looks and sounds. Different settings decide what it says. Both matter. They're different controls.

These sit next to the tutor's name and a short instructions field. You edit all of it the way you built the tutor: by chatting. Say "make the feedback gentler" or "less formal, more direct" and the change takes effect right away, no settings panel to hunt through.

The instructions field is also where you set limits. Tell the tutor what it will and won't take on: topics that are off the table, questions it should hand back to you, and the fact that it answers from your uploaded material rather than whatever it half-remembers. If you teach something regulated, this is where you make it decline out-of-scope advice instead of guessing. For an AI version of you facing your audience, that guardrail is what protects your name.

Tip: don't over-specify every field at once. Set the two or three that matter most for your subject, teaching style and feedback style are usually the ones learners feel first, and adjust the rest once you've previewed a real lesson.

Step 2: Generate or pick an avatar

Your tutor needs a face. You have two ways to get one: describe it and let Alltutors.ai generate a portrait, or pick from a set of preset characters if you'd rather skip straight to a finished look. Neither is more "official" than the other, plenty of published tutors run on a preset.

If you generate one, describe it the same way you'd describe anything else in chat: the vibe, the setting, the kind of presence that fits the subject you teach. You'll get a draft back to react to, not a final answer, so it's fine to ask for another pass if the first one doesn't land.

There's no wrong order here either. Some creators grab a preset first because it's fast, publish with that, and only come back to generate something more specific once real learners are going through the tutor. Others start with a precise mental picture and generate straight away. Either way you end up with a portrait that follows the tutor everywhere: dashboard, preview, and the public page a learner sees before their first lesson.

Step 3: Pick a voice

Pick a voice from the set on offer. These are curated synthetic voices, so none of them is literally your voice. Your job is to find the one closest to how you sound when you teach. You get a short sample. Listen before you commit. A voice that reads fine as a description can sound completely wrong the moment it's reading your real lesson content out loud.

Match it to your own delivery and pace, not to what feels "impressive." A coding tutor walking a nervous beginner through their first bug wants something calmer and slower than an IELTS coach drilling speaking practice against the clock. Pacing and warmth aren't just taste. They move comprehension and drop-off for anxious or non-native learners, which is why you judge the voice in preview against how you actually sound on camera.

Step 4: Preview it like a learner would

Preview it before you publish. This is the only step that shows the persona in motion. Open the preview and go through your own tutor the way a learner will: read a lesson, listen to the voice deliver it, get a question wrong on purpose and watch how the feedback lands. A persona reads fine on paper and can fall apart out loud. The preview is the only real test.

If something's off, tone too stiff, voice pacing wrong for the material, go back to chat and adjust it. The preview and the editing conversation stay in sync, so you're never stuck guessing whether a change took effect.

Step 5: Keep it consistent with how you teach

Persona settings are structured choices so learner one and learner one hundred get the same tutor. Our comparison of an AI tutor to a custom GPT goes deeper on why that consistency is the whole point.

Consistency isn't only about polish. A tutor whose manner keeps people engaged is how you fight the completion cliff, where most students quit a course long before the end. Bad reviews and refunds follow from that cliff, so the manner is doing real work.

Keep the persona close to how you'd teach this in person. If your course promises a warm, patient tutor but the feedback reads clipped and businesslike, that gap shows, and learners notice even when they can't say why. A better prompt won't fix it. Pick settings that match how you actually teach, then check them in the preview before anyone else sees them.

Building as a team rather than a single tutor? A shared workspace with roles and seats lets several creators work side by side. Our team workspace page covers what's included.

What's next

Haven't built your first tutor yet? Our quickstart guide walks the whole path, from a single prompt to a published, shareable tutor, persona and study plan together. Once the voice and face are right, open Alltutors.ai and go build it.