VEED Dubbing Pro: dialects and a voice-match slider

Dubbing has a reputation for sounding like a bad radio drama. Part of that reputation is deserved, and part is a set of habits that stuck around after the tools improved. VEED Dubbing Pro is the more controllable of the two VEED dubbing models: it translates the speech of a video, clones the voices, matches the lips, and gives you two things the standard version does not, a choice of regional dialects and a slider for how closely the new voice follows the old one. Here are three myths worth dropping, and the settings that go with them.

Myth 1: "Spanish is Spanish"

A viewer in Mexico City and a viewer in Madrid hear different words for the same thing, and both notice when the video was made for the other one. The target language field of the Pro model accepts regional tags, such as Mexican, Argentine, Chilean or European Spanish, Brazilian or European Portuguese, Canadian French, several kinds of English and Egyptian Arabic. Pick the one that matches the people who will watch. A base language such as plain Spanish still works; the region is optional. If you give a tag for a language that has no dialects, the region is ignored, and an unsupported dialect is rejected instead of silently guessed.

Myth 2: "The cloned voice should match perfectly"

This is the slider. Voice cloning strength runs from 0 to 10 and starts at 7. Higher keeps more of each speaker's voice and accent; lower sounds more natural in the new language. Perfect matching sounds odd, because the original voice carries the accent of the original language, and a Spanish line with an English accent is faithful and strange at the same time. Think of it as a dial between faithful and fluent.

• Try the default first. It is a good middle.
• Go up for a known face. A founder, a teacher or a streamer whose voice the audience recognises should still sound like themselves.
• Go down for anonymous narration. A tutorial voice-over has no personal voice to keep, so a natural-sounding line wins.
• Change one thing at a time, and keep the same thirty seconds for every try, so you hear the slider and not the clip.

Myth 3: "It translates everything"

The model translates speech, and it translates it literally enough to wreck a product name. The key terms field takes up to ten names and terms to keep, each short. Put the brand, the product and any phrase you want unchanged there before you run. Leave the source language on auto unless the video has a mix of languages or a heavy accent, in which case say what is spoken.

A simple order of work

1. 01Cut a thirty second piece where the speaker's face is clear.
2. 02Pick the dialect for your audience and fill in the key terms.
3. 03Run it at the default slider, then once higher and once lower.
4. 04Ask a native speaker which of the three sounds right, not which sounds most like the original.
5. 05Dub the full video, up to ten minutes per file, with the winning setting.

Where it still stops

Pro costs more than the standard model, and it is only worth it when the dialect or the voice match matters to your audience; for a quick internal video, VEED Dubbing is enough. Neither model fixes noisy source audio, overlapping speakers or a script that is already confusing, so clean the sound first, and cut the video into scenes if people talk over each other. A dub is also not a legal translation: for anything where each word counts, have a native speaker check it. More video and voice models are listed on the models page; to try this one, open the studio.