Videos

AI voice options

There are two different voice pickers here and they work at different scales. When a video is generated, the director chooses from eight characters, four styles across two genders. In the builder afterwards, every voice Google offers is searchable per scene. Most people only need the first.

Create a video with OneCraftThe four styles, how one is chosen for you, and the rule that overrides the choice

The product, format, length and language settings in the video generator, highlighted — Format, length and language are set here, not written into the brief.

The four styles

Energetic is the ad read, bright and forward. Instructional is measured and clear, made for a walkthrough. Warm is softer and closer, for anything sensitive or premium. Neutral is even and unmarked, which is the right choice when the words should carry and the delivery should not. Each has a male and a female character, so eight in total, and they are Google Chirp3-HD voices resolved to the locale of the language you asked for.

It is chosen from the mood, not the format

The director is told explicitly to pick the style and gender from the mood of your particular brief rather than defaulting by video type, and the instruction spells out the case that catches people: a calm, premium or sensitive subject wants warm or neutral even when it is a promotional video. So describing the tone in your brief is the single most effective thing you can do to influence the voice.

One hard override

A tutorial can never get the energetic voice. If the director asks for it in tutorial mode, the selector replaces it with instructional before the voice is resolved. That is a rule in code rather than a preference, and the reason is simple: ad delivery over a step by step instruction reads as a sales demonstration, and a viewer following along on their own screen stops trusting it.

The full catalogue in the builder

Once the video exists, each scene's voice dialog fetches the live Google voice list and lets you search it by language and by voice, which is far too large a set for a plain dropdown. So the eight director voices are the automatic choice, not the ceiling. Alongside the voice you set the narration text, a start delay in seconds, a playback speed of 0.85, 0.9, 1, 1.1 or 1.25 times, and whether that scene captions itself.

How it works, in three steps

Step 1
Describe the tone in the brief
Gentle, confident, matter of fact, urgent. That sentence is what the style and gender are picked from, and it works better than naming a voice.
Step 2
Listen to the first scene
If the read is wrong, it is wrong everywhere, because one voice carries the whole video. Fix it before editing anything else.
Step 3
Change it per scene if you must
Open the voice dialog, search the catalogue, pick a voice and regenerate. Each regeneration spends one credit.

The full walkthrough with screenshots is in the guide Add voiceover, captions and music to a video.

Limits worth knowing

Each voiceover clip generated in the builder spends one credit, so re-voicing a twelve scene video is twelve credits.
There is no emphasis markup, no pause tags and no pronunciation dictionary. Delay, speed and rewriting the sentence are the controls.
Two locales have no premium Chirp3-HD tier and fall back to an older Wavenet voice: Malay and Filipino.
Russian offers only a subset of the characters, so five of the eight are swapped for the two that exist there.

Questions people ask

Can I clone my own voice?

No. Voices are Google's synthetic set. There is no voice cloning, no upload of a voice sample and no custom voice training, so the narration is always a synthesised character rather than a person.

Will the same voice carry across a UGC video?

Yes, and that is the point of it there. One voice speaks every line and the same generated portrait is animated for each, so identity and voice cannot drift between cuts.

How do I get a slower read?

Drop the playback speed to 0.9 or 0.85 in the voice dialog. The scene retimes itself to the longer audio, and the original duration is remembered so putting it back shrinks the scene again.

Which style for an explainer?

Neutral or instructional. An explainer is asking somebody to follow an argument, and an energetic read makes every sentence sound equally important, which is exactly wrong when one of them is the point.

Make your own video

The button opens the generator with this use case already described. Change the wording to match yours, generate, then edit anything you like.

Create a video with OneCraft

AI video generator

You write a paragraph about what you are promoting. What comes back is not a storyboard or a set of stock clips to assemble, but a video that already plays: scenes picked and ordered, copy written into every box on screen, a narration track spoken aloud and captions running under it.

AI video script generator

A video script that reads well on a page usually plays badly. The reason is that the page has no timing. Here the script is produced scene by scene, and every line arrives attached to a duration, an on-screen headline and a layout that has already been chosen.

Slideshow video with narration

The reason most photo slideshows are unwatchable is that every picture gets the same three seconds and one voiceover runs across the lot. The interesting shot flies past, the filler lingers, and the narration is talking about image two while image five is on screen.

More finished work of this kind is on the video examples hub.