AI video ad guidesPet ads

How to lip sync an AI avatar to a script for an ad

Lip sync follows the voiceover, not the text. Paste your script and pick a voice, or upload a clean recording, then render a short test and check the mouth before you build the full ad. Fix problem lines one at a time and cut to product footage after the hook.

Updated November 20, 20267 steps5 prompts3,228 real ads

Make my video ads →

$1 today. Cancel anytime before day 7.

The steps
  1. 1Choose avatar and vertical formatIn HeyGen, open Single Scene, switch to Presenter Mode and pick your avatar. Set portrait for a 9:16 ad. If you start from a still image, create a Photo Avatar first.
  2. 2Write the script for speechKeep the opening line under ten words and write the way people talk. Spell out product names the way they sound.
  3. 3Paste script or upload voiceoverPaste the script and choose a voice, or upload one clean recording with a single speaker and no music. Play the audio preview and fix pauses and mispronunciations.
  4. 4Render a short test clipGenerate only the hook or the line with your product name. The audio preview cannot show lip sync, so watch the rendered clip.
  5. 5Regenerate weak lines individuallyIf the mouth looks off, redo that line with a cleaner or slightly slower read. Do not shift the whole audio track.
  6. 6Cut to product footageKeep the avatar on camera for the hook and the call to action. Show the product during the demo while the voiceover keeps running.
  7. 7Add captions and music lastLock the dialogue timing first. Then add captions, music and the end card in your editor.
On this page
  1. The steps
  2. Real pet video ads
  3. Decide where the voice comes from
  4. Write the script for the mouth
  5. What long-running pet ads teach about the opening line
  6. Set up the avatar scene in HeyGen
  7. Test, judge and fix the lip sync
  8. Assemble the ad in short beats
  9. Prompts to copy
  10. Which voice input fits your ad
  11. Mistakes to avoid
  12. Questions

Real pet video ads to model yours on

3,228 video ads from 395 advertisers in the Meta Ad Library, read October 5, 2026. The median one ran 44 days, and an ad still paid for after months is the closest public sign that it works. Live ads first.

  • Ran 8 days
  • Ran 4 days
  • Ran 4 days
  • Ran 4 days
  • Ran 18 days
  • Ran 8 days
  • Ran 12 days
  • +3,216 more pet video ads

    Hooks, landing pages and run length for each

  • Ran 15 days
  • Ran 18 days
  • Ran 3 days
  • Ran 31 days

Every creative, with its copy, format and run length

The openers that ran longest

  • ShawnCo Essential Dog Harness!Shawnco · running 1,024 days
  • {{product.name}}Whunde · running 953 days
  • {{product.name}}ACANA Italia Petfoods · running 921 days
  • Permanently Removes Pet Stains and Odors!FurryFreshness · running 800 days

Decide where the voice comes from

Lip sync follows audio, not text. The avatar tool turns your script into speech, then animates the mouth to match that speech. So the thing you are really controlling is the voiceover. Decide early whether the tool generates it or you upload it, because that choice sets how much control you have over timing, emphasis and how your brand name is said.

Pasting a script is the fastest route. You pick a voice, paste your lines and the tool does the rest. It suits short product ads where speed matters more than performance. Upload a recording when pacing matters. A recorded read lets you place a pause before the offer, lean on a key word and say a brand name the way your customers say it.

Whatever you choose, keep the audio clean. Use one speaker, no background music and no room echo. Music and second voices make the mouth animation harder to judge and harder to fix. Add music only in the final edit, once the dialogue timing is locked and approved.

  • Pasted script: fastest, good for short product ads
  • Uploaded recording: best for pacing and brand pronunciation
  • Always one speaker, no music under the dialogue

Write the script for the mouth

A script that reads well on a page can look strange on an avatar. Long sentences force the voice to rush or to pause in odd places, and the mouth makes both obvious. Write short lines. One idea per sentence. Plan on roughly two to three spoken words per second, so a 20 second ad carries about 40 to 60 words of avatar speech, not 100.

Open with a line you can say in under three seconds. Eight words or fewer is a good limit. That line is also your first lip-sync test, so a short hook keeps each render cheap in time and easy to judge. Put the product name in the second or third beat, not the first, so a mispronunciation does not ruin the opening.

Fix pronunciation in the text before you generate. If the voice says your brand name wrong, try a phonetic spelling, add a hyphen between syllables or split the name with a space in the script only. Keep the on-screen caption spelled correctly. Commas make short pauses and full stops make longer ones, so use punctuation to shape the rhythm.

  • Hook: 8 words or fewer
  • 20 second ad: about 40 to 60 spoken words
  • Fix brand name spellings in the script, not in the captions
  • Use commas and full stops to control pauses

Want pet video ads like these for your product?

The pet video ads that kept running, read from the ad libraries. Every change waits for your approval.

What long-running pet ads teach about the opening line

The Meta Ad Library shows 3,228 pet video ads from 372 advertisers, with a median run of 44 days. The ads that ran far longer than that share one habit: the opening line makes a single promise you can say in one breath. FurryFreshness ran "Permanently Removes Pet Stains and Odors!" for 800 days. WINPRO Pet ran "See How Plasma Can Help Your Dog" for 433 days. JustFoodForDogs ran "Your Dogs New Favorite Food" for 342 days.

Notice what is missing. None of these lines introduce the brand, tell a backstory or warm up. Each states a result or a reason to keep watching, in about five to seven words. That is exactly the line an avatar can deliver cleanly, because short phrases give the lip sync less room to drift. Petoir ran a warmer opener, "Let your furry friend live their best, healthy life!", for 774 days, so a friendly tone works too, as long as the promise is clear.

Use the same pattern for your avatar hook. Pick one outcome, say it in under ten words and look straight at the camera. Then show the product doing the thing. The Sprites ads library shows real ads from a category, so you can scan what your competitors open with before you write your own line. For more on the format, see pet video ads.

Set up the avatar scene in HeyGen

Open Single Scene and switch to Presenter Mode, then choose your avatar. Set the canvas to portrait so the clip fits a 9:16 placement without cropping. If your starting point is a still image, make a Photo Avatar first. Do not drop the image in as a regular asset, because a regular asset will not be animated to speak.

Next, add the audio. Paste the script and choose a voice, or upload your recording. Play the audio preview and listen for awkward pauses, wrong stress and mispronounced product names. The preview checks the voice only. It cannot show you the mouth. Synthesia follows the same basic logic: you write in the script box, pick an avatar and voice, and the video renders from that text.

Pick the framing before you generate. A chest-up shot with the face centered gives the mouth the most pixels, which makes sync problems easier to spot and the finished ad easier to read on a phone. Keep the background plain or simple, since you will cut away from the avatar during the demo anyway. If you want a more casual feel, read UGC ads for how creator-style shots are built.

Test, judge and fix the lip sync

Never render the full ad first. Generate the hook, or the line that contains your product name, and watch the finished clip on a phone-sized screen with sound on. Check three things: does the mouth open and close on the same beats as the words, does the delivery sound natural, and does the face stay still between sentences. Judge it at full speed, then once more with the sound off.

When a line looks wrong, regenerate only that line. Slow the read slightly, shorten the sentence or swap one hard word for an easier one. Moving the whole audio track rarely helps, because the problem is usually in one phrase. If you uploaded a recording, re-record that line with a cleaner read and replace just that piece.

Keep a short log of what worked. Note which voice, pace and punctuation gave a clean result, and reuse them across the rest of your ad set. Consistent settings mean your avatar sounds like the same presenter in every variant, which matters when you test several hooks against each other.

  • Render the hook first, not the whole ad
  • Watch at phone size with sound on, then muted
  • Redo one line at a time
  • Reuse the voice and pace that tested clean

Assemble the ad in short beats

Use three beats: hook, demo or benefit, call to action. Put the avatar on camera for the hook, about the first three seconds, and again for the call to action in the last three to five seconds. In the middle, cut to product footage while the voiceover keeps playing. This hides any imperfect sync, keeps the viewer's eyes on the product and stops the avatar from feeling like a talking head for the whole ad.

Add captions after the dialogue timing is locked, keeping them inside the safe zone so the platform buttons do not cover them. Then add music at a low level under the voice. Export at 1080 by 1920. Check the platform rules in Meta ad specs before you upload, and match the button to the offer. The most common pet buttons are Shop now, Learn more, See details and Get offer.

If you need a steady stream of variants, change one thing at a time: the hook line, the avatar or the demo clip. Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. You can read more at Director.

Prompts to copy

Paste each one as written, then change the product, the setting and the light to match your own footage.

  1. Hook: stain remover promise

    Vertical 9:16 presenter shot, a friendly adult avatar in a bright living room, chest-up framing, looking straight into the camera with a relaxed smile. Script: "Pet stains gone in one wipe." Natural pace, one beat of silence after the line. Hands stay below the frame. Soft daylight, plain background, no logos, no on-screen text.
  2. Hook: dog food result

    Vertical 9:16, a warm kitchen background slightly blurred, an avatar presenter centered at chest level, eyes on camera, upbeat but calm. Script: "Your dog will not stop licking the bowl." Slightly slower read, clear stress on the word bowl. Keep the face still between sentences. No brand names, no captions in the render.
  3. Product name test line

    Vertical 9:16, avatar presenter in a neutral studio, chest-up, steady eye contact. Script: "Meet Pawra, the chew that lasts." Pronounce the brand as PAW-rah, short pause after the comma. Use this as a pronunciation and lip-sync test only. Render a single short clip with a plain background and no music.
  4. Demo: voiceover over product

    Vertical 9:16 product close-up, a generic dog treat pouch on a wooden table, a hand pours treats into a bowl, shallow depth of field, soft side light. No people visible, no logos. Leave the clip silent so the avatar voiceover can run over it. Slow push-in over about six seconds.
  5. CTA: avatar returns on camera

    Vertical 9:16, same avatar and background as the hook, chest-up, a small confident nod at the end. Script: "Tap shop now and try it today." Steady pace, slight emphasis on shop now, hold one second of silence after the final word. Plain background, no logos, no text in the render.

Which voice input fits your ad

InputBest forWatch out for
Pasted script with tool voiceFast tests and short product adsBrand names and pauses may need respelling
Uploaded voiceoverPrecise timing, emphasis and pronunciationNeeds one clean speaker and no background music
Speech to Speech (voice mirroring)Keeping your own delivery on the avatarYour recording sets the pacing, so re-record weak lines
Photo Avatar from a stillStarting from one imageCreate it as a Photo Avatar, not a regular asset
Synthesia script boxWriting the script directly in the editorSame rule: test the render, not just the audio

Mistakes to avoid

  • Judging lip sync from the audio preview instead of rendering a clip.
  • Writing long sentences that force the voice to rush or pause oddly.
  • Uploading a recording with music or a second voice under it.
  • Leaving a mispronounced brand name in the script because the caption looks right.
  • Keeping the avatar on screen for the whole ad instead of cutting to the product.
  • Shifting the entire audio track to fix one bad line.

Rather have the videos made for you?

Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. You never write a prompt.

Apply to Director by Sprites →

Questions

Does the avatar lip sync to my text or to the audio?

It syncs to the audio. If you paste a script, the tool turns it into speech and animates the mouth to match. If you upload a recording, it animates to your recording.

Should I paste a script or upload my own voiceover?

Paste a script for speed. Upload a voiceover when you need exact pacing, emphasis or brand pronunciation. Use one speaker and no background music.

Why does the lip sync look off on my avatar?

The most common causes are long sentences, awkward pauses, a mispronounced word or a noisy recording. Regenerate only the weak line with a cleaner, slightly slower read.

Can I check lip sync before rendering the whole video?

The audio preview checks the voice only. To see the mouth timing you need to render the clip, so render a short test of the hook or product line first.

Can I make a vertical AI avatar ad from a photo?

Yes. Create a Photo Avatar from your image first, then use it in Single Scene with a portrait format. Do not add the photo as a regular asset.

How long should the avatar talk in a short ad?

Keep the avatar on camera for the hook and the call to action, about three seconds each. Use product footage in the middle while the voiceover continues.

Keep going

Get pet video ads made for your brand

  • The pet video ads that kept running, read from the ad libraries
  • Finished, on-brand video ads, reviewed by a human
  • Every change waits for your approval