Give your video ads a voice

A product video that talks sells better than a silent one, and you don't have to record anything. A prompt writes a short voice script for every product from its own details, and a Studio video template reads that script aloud when it renders the product's video. Each product gets its own pitch, in a voice you pick once for the whole template.

Two steps, both set up once: a prompt that writes the scripts, and an audio element in your video template that speaks them.

Step 1: Write the voice scripts with a prompt#

Open the project, go to its Creative tab and click Create prompt. Pick Write text-to-voice scripts as the type of prompt. It starts with instructions written for spoken copy and produces content of the Voice script type, which is what the template looks for later.

Adjust the instructions the way you would for any text prompt, with the ear in mind rather than the eye:

• Short. Two or three sentences is plenty for a 6-second video. Say how long you want it in the instructions ("at most 25 words").
• Spoken, not written. Ask for plain words and full stops. No bullet lists, no "click here", no prices written with symbols the voice would read out oddly.
• One reason to buy. Name the product, give one concrete reason, end with the price or the offer if you want it spoken.
• The catalog's language. The script is spoken in the language it is written in, so a prompt scoped to a Swedish catalog writes Swedish scripts. A product sold in three markets gets three scripts, one per catalog.

Publish the prompt. Like every prompt it runs for every product in scope, and again whenever a product changes, so new products get a script as they arrive. The scripts show up on each product's Content tab as Voice script content, where you can read, approve or reject them like any other AI copy. Writing them spends credits as Prompt-written text.

Start with a small product set while you tune the wording: a published prompt writes for everything in its scope, and the voice-over is only as good as the script.

Step 2: Read the script aloud in the video template#

Open the video template in Studio, or create one. If the template has no duration yet, turn on Video in its Settings tab (see Video templates).

1. Add an audio element with the Audio tool. It has no visual presence: it lives in the Elements tree and on the Timeline.
2. In its Binding section, add a Voice script placeholder. The element's source now reads bound to product data, and the upload controls disappear: there is no file, the script is spoken at render time.
3. Pick the voice. Click the voice icon at the left of the binding field, or the Voice: line in the Audio section, to open the Voice dialog:
   • Automatic Google voice is what every audio element starts with: a Google voice in each product's catalog language, with the Speed you set. Nothing changes it unless you pick another source here.
   • Google voice by language lets you choose one language and one voice for every product.
   • One of your ElevenLabs accounts lets you pick from the voices in that account, the ones you cloned there included, with a Model, a Speed and the Advanced settings. The account is connected on the Integrations page.
4. Click Save, then preview a product. The Audio section names the voice, for example "Voice: Swedish / Female / Wavenet C · Google", and Play beside it plays that product's voice-over. Switch the preview product and you hear its script instead.

Place the audio element on the Timeline where the speech should start, and give the video enough duration for the longest script. Background music is a separate audio element with an uploaded file; the two play together.

The full set of voice settings, and what happens when a voice can't speak, is in Video templates.

Step 3: Render it for every product#

Nothing more is needed on the channel side. Attach the template to a channel's video field as described in Generate a video ad from your products, publish, and every product's video renders with its own script spoken in the voice you chose. A product whose script has not been written yet renders once it has; a product the prompt skips, because it lacks what the prompt reads, has no voice-over and the video is missing content until it does.

Using your own voice#

A founder reading the pitch, or a creator you work with, is the version customers remember. That voice lives in your own ElevenLabs account: clone it there, give the consent ElevenLabs asks for, then connect the account on the Integrations page and pick the voice in the Voice dialog. Renable stores no recording and no sample, only the connection. A voice ElevenLabs still has to verify is dimmed in the list until you finish the check in ElevenLabs.

Voice-overs spoken through an ElevenLabs account use that account's characters, billed by ElevenLabs. That includes the one Studio makes when you preview a product after changing the voice. A voice-over already made for the same script, voice and settings is reused, so a re-render of an unchanged product spends nothing.

Or ask Aimée#

On the project's Creative tab, ask Aimée to create the voice script prompt and tell her the tone you want. In Studio, ask her to add the audio element with its Voice script binding, to list the voices you can use, and to set one: "use Anna from our ElevenLabs account", "make it a little slower", or "go back to the Automatic Google voice". She can't record or upload audio, and she can't clone a voice: that happens in ElevenLabs.

FAQ#

The preview says the voice-over couldn't be made just now.

The voice service didn't answer in time. Change the preview product or save the voice again and it is tried once more. For an ElevenLabs voice, check that the account still has characters left.

The audio element shows a warning instead of a voice-over.

Hover the warning icon on the binding field: it says why. The usual reasons are a product without a Voice script yet (the prompt hasn't run, or skipped it), an ElevenLabs account that was disconnected, a key ElevenLabs no longer accepts, or a cloned voice still awaiting verification in ElevenLabs.

Can I control the speed from the script?

No. Speed is a voice setting in the Voice dialog, the same for every product. A [sr=...] token at the start of a script is ignored.

Does every product need its own script?

Yes, if you want a voice-over for it. A product the prompt skipped has no script and no voice-over. If you'd rather speak a fixed line for every product, bind the audio element to text you write yourself instead of the Voice script placeholder.