Use this guide to direct Narration Box voices with style instructions of all kinds.

This is a creative guide for making AI narration sound more human, intentional, cinematic, expressive, and context-aware.

Style instructions help you tell the voice how to perform a paragraph or block of text. They can guide emotion, energy, pacing, attitude, vocal texture, delivery style, intensity, and character presence.

The best results come from experimenting. Users are encouraged to experiment with emotions and expressions and see what works and test the limits. This guide presents a description of all the possibilities we have seen and tested. Give this guide a go to see what is currently working to spark your curiosity.

A single word can work.

A short descriptive phrase can work.

A more specific performance direction can also work.

The goal is to try different versions until the voice matches the scene, audience, and format.

Style instructions can help you shape:

The same text can feel completely different depending on the instruction.

Example:

Text: “I didn’t expect you to come back.”

Possible style instructions:

Each version creates a different character, scene, and meaning.

You can go for long prompts that describe emotions in detail or short. If your prompt can be diluted to one or two words, you are advised to shorten but our voices also work on super specific prompts, for example:

A single word also works:

Short descriptive phrases work too:

You can also be highly specific when the performance needs a particular identity, region, accent, format, character type, or scene style.

For example, a style instruction can describe:

This means users are not limited to simple emotions. They can experiment with more detailed voice direction when they want a very specific result.

Useful prompt patterns:

More specific instructions are useful when the scene needs nuance:

The best workflow is:

Sometimes a specific narrator style or accent direction gives the voice the exact character the project needs or a very niche specific or situational description of the emotion.

There are two ways to direct performance:

This controls the overall delivery of the paragraph or text block.

Example style instruction:

calm and reassuring

Text:

You are safe now. Take a deep breath. We are going to handle this one step at a time.

This is useful when the whole paragraph should carry the same emotional tone.

Inline tags are used inside the text when you want a specific moment to shift emotion, pacing, or texture.

Example:

I thought I was ready for this, [nervous] but the moment the door opened, I forgot how to breathe.

This is useful when the sentence changes emotional direction.

If a text block has both a style instruction and an inline tag, the voice will usually follow the inline tag at that specific moment.

Example style instruction:

calm

Text:

Everything looked normal at first. [whispering] Then I heard someone breathing behind the wall.

The block begins with the calm instruction, then the inline tag shapes the specific moment.

Inline tags work best when they are placed after the first word of the sentence or block.

I [whispering] knew someone was standing behind me.

The room [eerie] felt colder than it had a minute ago.

She [angry] told him never to come back.

Avoid:

[whispering] I knew someone was standing behind me.

When the inline tag is placed right at the beginning of the block, the narrator may read it as part of the text instead of treating it as a performance direction.

Write at least one normal word first, then add the inline tag.

For best results, use English tags and English style instructions even when the transcript is in another language.

Example Hindi text:

Mujhe [nervous] lag raha tha ki koi mere peeche khada hai.

Example Spanish text:

No sabía qué decir, pero [soft] quería que ella se quedara.

Example French text:

Il faisait sombre dans la pièce, et [whispering] quelqu’un respirait près de la porte.

The spoken content can be in another language, but the performance direction should remain in English.

AI voices are context-aware. This means the voice does not only read the style instruction. It also reacts to the sentence itself.

For example, the instruction “happy” may sound different depending on the sentence:

I finally got the job.

You came back.

We won. We actually won.

I can’t believe this is happening.

The same style instruction can produce excitement, relief, disbelief, warmth, or joy depending on the words.

That is why testing matters.

A good prompt is not always the longest prompt. Often, the best instruction is the one that gives the voice just enough direction without over-controlling the performance.

For every emotion or delivery style, test three versions:

sad

quietly sad

trying not to cry while speaking softly

Then compare which one fits the scene best.

For example:

Text:

I kept your letter all these years. I don’t know why. Maybe I was waiting for a reason to let go.

Prompt options:

Each one may work, but each one creates a different version of the scene.

Keep style instructions short and clear.

Good:

Too much:

Better:

This section gives users copyable style instructions grouped by emotional family. Use these as starting points, not fixed rules. The same prompt can behave slightly differently depending on the voice, the sentence, the language, and the emotional context of the paragraph.

For best results, test 2–4 nearby prompts before choosing the final one. You can definitely make your own prompt and test with different words and emotions.

Use these when the scene needs joy, charm, celebration, relief, playfulness, optimism, or light emotional lift.

Style instructions:

Voice textures:

Example:

I [excited but controlled] can’t believe we actually did it. After all those late nights, all those failed attempts, this is finally real.

Good for:

Be careful with:

Use these when the scene needs grief, vulnerability, disappointment, regret, loneliness, memory, emotional heaviness, or quiet pain.

Style instructions:

Voice textures:

Example:

I [fragile] thought I had made peace with it. But when I saw the empty chair, everything came back at once.

Good for:

Be careful with:

Use these when the scene needs irritation, rage, bitterness, confrontation, disgust, impatience, threat, or controlled intensity.

Style instructions:

Voice textures:

Example:

You [cold] knew exactly what would happen, and you still let me walk into that room alone.

Good for:

Be careful with:

Use these when the scene needs panic, nervousness, dread, unease, hesitation, shock, paranoia, or desperation.

Style instructions:

Voice textures:

Example:

I [trembling] tried to call out, but my voice broke before the words could leave my mouth.

Good for:

Be careful with:

Use these when the scene needs tenderness, attraction, warmth, affection, longing, comfort, softness, or emotional closeness.

Style instructions:

Voice textures:

Example:

I [soft smile] don’t know when it happened. I just know that every quiet moment started to feel better when you were there.

Good for:

Be careful with:

Use these when the voice needs to sound clear, firm, persuasive, polished, grounded, professional, heroic, or leader-like.

Style instructions:

Voice textures:

Example:

We [calm authority] do not need to move faster. We need to move with discipline, clarity, and intent.

Good for:

Be careful with:

Use these when the scene needs dread, darkness, mystery, danger, suspense, evil presence, or cinematic horror.

Style instructions:

Voice textures:

Example:

The hallway [eerie] stretched farther than it should have. At the end of it, something smiled in the dark.

Good for:

Be careful with:

Use these when the voice needs to feel peaceful, centered, slow, gentle, reassuring, or emotionally steady.

Style instructions:

Example:

Take a slow breath in. Let your shoulders soften. For the next few moments, there is nowhere else you need to be.

Good for:

Be careful with:

Use style instructions to guide narration, character emotion, chapter tone, and scene intensity.

Recommended prompts:

Where it works well:

Where to be careful:

Use style instructions to increase retention, energy, and clarity.

Recommended prompts:

Where it works well:

Where to be careful:

Use style instructions to create natural host-like delivery, relaxed conversation, and expressive storytelling.

Recommended prompts:

Where it works well:

Where to be careful:

Use style instructions to make learning content clear, patient, and easier to follow.

Recommended prompts:

Where it works well:

Where to be careful:

Use style instructions to make the voice persuasive, clean, benefit-led, and brand-safe.

Recommended prompts:

Where it works well:

Where to be careful:

Bad:

I [happy] opened the door. I [excited] saw the box. I [shocked] picked it up. I [curious] looked inside.

Better:

Style instruction:

excited and curious

Text:

I opened the door and saw the box waiting there. For a second, I just stared at it. Then I picked it up, wondering what could possibly be inside.

Avoid:

happy, sad, angry, romantic, nervous, calm, excited, sarcastic

Better:

conflicted

nervous but hopeful

angry but trying to stay calm

The voice can suggest laughter, fear, breathiness, tension, whispering, softness, or intensity.

But it cannot always produce every long and very extreme physical sound. But it can in some cases, you are encouraged to test the limits of the voices.

For example, prompts like these may be inconsistent but can also produce some really creative results:

Use style instructions for performance direction, not full sound design.

Vague:

make it better

more human

cinematic

Better:

low and tense

soft and reflective

energetic

serious and documentary-style

warm and reassuring

Results may vary depending on:

For example, a sentence about a calm product tutorial may just not become convincing horror narration just because the style says “terrified.” The text and the style instruction should support each other.

Good voice direction works best when the writing, prompt, and voice choice are aligned.

The text itself matters.

Weak pairing:

Style instruction:

terrified

Text:

Today we will learn how to export an audio file.

Better pairing:

Style instruction:

calm instructor

Text:

Today we will learn how to export an audio file step by step.

Strong pairing:

Style instruction:

terrified

Text:

I heard the lock turn from the other side, even though I was alone in the house.

The prompt gives direction, but the sentence gives the voice something to perform.

Not every voice will respond the same way to every instruction.

Some voices are naturally better for:

A prompt can shape the voice, but the base voice still matters.

For best results:

I walked into the room and saw the envelope on the table. Nobody had touched it. Nobody had even come near it. But somehow, it was open.

suspenseful

quiet and tense

I walked into the room and saw the envelope on the table. Nobody had touched it. Nobody had even come near it. But somehow, it [eerie] was open.

Style instruction:

low and suspenseful

Text:

I walked into the room and saw the envelope on the table. Nobody had touched it. Nobody had even come near it. But somehow, it [eerie] was open.

This shows how style instruction and inline tags can work together without overloading the paragraph.

Before generating a full chapter, course, podcast, or video script:

This avoids wasting time on long generations that do not match the intended style.

Narration Box voices are not limited to flat text-to-speech.

With style instructions and inline tags, creators can build:

The real power is not just generating a voice. It is directing a performance.

Style instructions can also be used to guide accent, region, and speaking style.

This is useful when your content needs a voice that feels more local, more character-specific, or better matched to the audience. You can ask for broad accents, regional accents, or a softer hint of an accent.

Examples of accent prompt patterns:

You can also combine accent with emotion or use case:

Accent prompting is experimental. Some voices will respond more strongly than others. Some voices may produce only a light regional flavor instead of a perfect accent. That is normal.

For best results:

Broad prompt:

British accent

More specific prompt:

soft Yorkshire accent, warm and conversational

Use accent directions when they help the listener believe the narrator, character, brand, or region. Do not use them just to make every voice sound different. The accent should support the content.

Style instructions can dramatically improve performance, but they are not magic buttons.

They cannot guarantee:

The best results come from pairing the right voice, the right writing, and the right instruction.

Now try a short sample with one voice, one paragraph, and three different style instructions.

Start simple. Then add texture. Then add one inline tag.

That is usually enough to turn plain narration into a directed performance.