One AI brand film, three languages: what needs to change?

I start a language version with the message, then decide where that message lives: in subtitles, an offscreen voice or a person speaking on camera. Translating those words is one part of the job. The new film also needs its own text layout, listening check and route to the right page. I would agree those pieces before asking for another export.

A small plant shop, one invitation

Imagine a fictional plant shop making a short film about bringing a little greenery home. A plant sits beside a watering can; the closing invitation is “Find a little green space”. I wrote this example for the article. The generated picture is an illustration, and the English, Russian and Spanish lines in the worksheet are teaching examples, not a tested campaign.

For Spanish I might write “Haz sitio para algo verde”. For Russian, “Добавьте немного зелени”. I am keeping the invitation gentle, rather than forcing the same number of words into each version. Before production, I would ask a fluent reviewer whether each line sounds natural for the intended customer and whether it still asks for the same next step.

Choose how the audience will receive the words

Subtitles can preserve the original performance while adding another reading language. A new offscreen voice can carry the story when the viewer is listening and the picture does not depend on visible speech. A person addressing the camera creates another task: the translated performance has to work with the face and the shot. I would compare those production scopes before choosing a tool.

For our plant film, there is no visible speaker to match. A translated text version or an offscreen voice may already serve the idea. If a shopkeeper later speaks in a close-up, I would test one short translated line before committing the whole film. A convincing mouth shape alone does not tell me that the meaning or pronunciation is right.

Three production routes for the same message

Keep the words on the pot separate from the voice

A translated voice does not replace writing already baked into a picture. If the plant tag contains a shop name or an offer, that layer needs its own decision. I would keep an editable text layer and a clean image wherever the planned design allows it. That gives the editor a direct way to change the words without asking a generator to redraw a real brand name.

The same applies to the final frame. A Spanish invitation that sends someone to an English-only form may create work for the visitor at the very moment they want to act. I would follow the destination on a phone, check the language actually available and agree the fallback when a matching page does not exist. A language label should describe the destination honestly.

• Write down the exact brand names that must stay unchanged
• List subtitles, speech, text in the picture and the final link separately
• Name the intended region and tone rather than using a flag as a language
• Keep an approved source script beside each language version

Give one short scene to a real reviewer

I would send a reviewer the picture and sound, the original intention and the translated line together. “Does this feel like a friendly invitation to visit the shop?” is a more useful question than “Is the translation correct?” A line can be grammatically sound and still sound stiff, overly familiar or unlike the person speaking.

Then I would watch the exported version from start to finish. Check that a name is pronounced as agreed, a quiet phrase remains audible and a caption has not survived from the wrong language. If a line changes after that review, mark its checks open again. Approval belongs to the version that was watched; it should not silently carry over to new words.

Ask what one language version includes

A useful request to SOLONSKY Films would name the source film, target languages, region, chosen speech route and where each version will appear. I would also ask who provides the translation, who reviews the language and what happens if the revised words need a different edit. Those answers define the work more clearly than a promise to “translate the video”.

YouTube documents separate additional audio tracks and translated titles and descriptions. Adding an audio track does not create the translation for you. I would check the features available in the actual channel and player before planning delivery. A downloadable MP4, a subtitle file and an alternate audio track serve different publishing workflows; agree which ones are needed.

Make the remaining decisions visible

Use the sheet above to choose a route and edit the invitation for each language. The picture shows where a short line might sit; it does not judge typography in a finished video. The checkboxes are your own review notes. Changing a phrase or route clears the checks for that language, while the other two versions stay in place.

Download the note if it helps you brief an editor. Everything in the exercise stays in this page; it does not translate text, generate a voice, assess lip sync or submit an approval. I would bring the three lines and the unresolved checks to the next conversation, then decide which sample needs making first. Reloading restores the teaching example.

Questions

Sources

• YouTube Help — multi-language features

  Official delivery documentation, checked 27 October 2026; channel access and player behaviour need checking separately

• Can your AI film tell its story with the sound off?

• Will you be able to change your AI film next month?

• Voiceover or dialogue: give your AI film the right voice

Watch an authored AI film

An authored film from SOLONSKY Films. It illustrates the storytelling approach.

Play The Impossible Brief

The video provider connects only when you press play. Watch on YouTube.

Have a film in mind? Discuss your film with me at SOLONSKY Films.