The Power of Custom Prompts: How to Control the Tone and Style of Your AI Translation
Stra.ai's translation engine is context-aware — it reads your full script before translating, which already puts it ahead of basic machine translation. But the custom prompt feature lets you go one step further and tell the AI exactly how the translation should sound. This guide explains how that works, what each setting actually does, and how to use them together to get output that fits your content from the first pass.
Two separate controls, two separate jobs
Before diving into prompts it helps to understand that Stra.ai has two distinct settings that affect how your content sounds — and they do completely different things.
The translation prompt controls the style of the written translation. It tells the AI how the words should be chosen formally or informally, with sports terminology or casual slang, in a news anchor register or a teenage conversational tone. This is about word choice and speech style, not formatting.
The voice model controls how the AI delivers those words once the translation is done. This only applies to AI Dubbing. It affects emotion, intonation, and energy whether the voice sounds calm or excited, flat or expressive.
They work at different stages of the same pipeline. The translation prompt shapes what gets written. The voice model shapes how it gets performed.
The translation prompt — controlling word choice and style
The translation prompt field appears in the Create popup for both AI Subtitles and AI Dubbing projects. It is optional , if you leave it blank, Stra.ai produces a neutral, accurate translation with no particular style applied.
When you fill it in, you are giving the AI a brief for how the translation should feel. There is no character limit. Write it the way you would brief a human translator.
The built-in style presets
If you do not want to write a custom prompt from scratch, Stra.ai offers five presets:
Live Sports Commentary — High energy, fast-paced, uses sports-specific vocabulary and commentator phrasing. Best for match highlights, sports recaps, and action content.
Casual Drama/Show Style — Natural conversational language, emotionally expressive, reads the way people actually talk. Best for entertainment content, vlogs, reality shows, and lifestyle videos.
Formal News Report — Neutral, authoritative, structured. Best for documentary content, corporate videos, educational explainers, and anything that needs to sound credible and professional.
Teen/Gen Z Style — Informal, current slang, relaxed register. Best for content aimed at younger audiences on social platforms.
Child Friendly Style — Simple vocabulary, clear sentence structure, warm and approachable. Best for kids' content, educational material for younger learners, and family-oriented videos.
Writing your own custom prompt
Select Custom Prompt from the dropdown and a text field opens. Write a brief describing the style you want. Think of it as a note to a translator, the more specific you are, the better the output matches your expectations.
A few examples of prompts that work well:
"Use casual, friendly language between two people in their mid-20s. Keep it conversational and natural, like texting a friend."
"Translate formally. This is a corporate training video , keep terminology professional and avoid colloquialisms."
"Use sports commentator language. High energy, present tense, short punchy sentences."
"The speaker is a chef explaining a recipe. Use cooking vocabulary naturally and keep the tone warm and instructional."
The key is to describe the relationship, the context, and the register , not just say "translate well." The more context you give the AI, the more targeted the word choices will be.
The voice model — controlling emotion and delivery (AI Dubbing only)
Once the translation is done, the voice model determines how the AI performs it. This setting only applies to AI Dubbing projects. There are two options:
ElevenLabs — Natural voice generation with voice cloning. Best when you want the dubbed voice to closely match the original speaker's tone and character. Good for personal content, vlogs, interviews, and anything where vocal identity matters.
Gemini TTS — Best for emotion and tone control. When you select Gemini TTS a separate field appears asking "How should the voice sound?" This is where you direct the AI's performance , not the words, just the delivery.
Examples of Gemini TTS voice directions:
"Speak with high energy, like a sports commentator calling a live match."
"Sound calm and authoritative, like a documentary narrator."
"Angry but trying to stay composed , tension underneath a controlled surface."
"Warm and encouraging, like a teacher explaining something to a nervous student."
Using both together
The most powerful setup is combining a translation prompt with a Gemini TTS voice direction , one shapes the words, the other shapes the performance.
For example, a sports highlights video might use:
Translation prompt: "Use sports commentator language. High energy, present tense, short punchy sentences."
Voice direction: "Sound like a commentator calling a live match , excited, urgent, building intensity."
A corporate training video might use:
Translation prompt: "Formal professional language. Avoid slang. Clear and instructional."
Voice direction: "Calm, measured, authoritative. Like a senior professional explaining a process."
When both settings are aligned the output feels intentional rather than generic.
You can adjust everything in the editor
If the first pass does not sound right, nothing is locked in. Once your project is processed you can edit the translation line by line in the editor, and adjust the voice direction per line or globally before regenerating audio.
The prompt you set during project creation is just the starting point , not the final word. Use the "generate translation first" toggle in the dubbing setup if you want to review and refine the translation before committing to voice generation at all.
What to do next
To set up your project and find the prompt field → A Guide to the Stra.ai Dashboard
To edit translation and voice direction after processing → Subtitle Editing Pro Tips
To fine-tune AI voice emotion in the dubbing editor → The Dubbing Work Area
To understand credit costs before you start → Token Economics Guide
→ Continue here: Token Economics — Understanding Credit Estimation and Pricing Plans
