# AI Avatar Video Generator > An AI avatar video generator turns a portrait plus an audio track into a talking video. Canonical: https://www.makeviral.ai/tools/ai-avatar-video-generator Updated: 2026-09-17 Published: 2026-09-17 Author: MakeViral Team (https://www.makeviral.ai/about) Publisher: MakeViral (https://www.makeviral.ai) An AI avatar video generator turns a portrait plus an audio track into a talking video. In MakeViral you write a script of 35 words or fewer, a voice reads it, and the creator portrait is lip-synced to that recording. One clip runs from 3 to 15 seconds at 1080 x 1920. Write up to 35 words. A voice reads them, the portrait is lip-synced to that audio, and a vertical talking clip comes back ready to upload. **Start now:** [Make a talking video](https://www.makeviral.ai/influencer) (opens the MakeViral app). Part of the [AI Influencer Generator](https://www.makeviral.ai/ai-influencer-generator). ## What is an AI avatar video generator? It is a lip-sync pipeline with three stages, and knowing the stages tells you where quality comes from and where the cost comes from. 1. **A portrait.** On the [AI creator surface](https://www.makeviral.ai/influencer) you cast a creator: a preset, a written description, or a face photo you upload. The portrait renders 9:16 and is saved with a name and a voice. The [AI UGC creator page](https://www.makeviral.ai/tools/ai-ugc-creator) covers casting properly. 2. **An audio track.** You type a script, or have one drafted from the product name, the description and the angle you pick. The creator's voice reads it. The audio is what you are billed on. 3. **The lip sync.** The portrait is animated to match that audio, so the mouth, the jaw and the head move with the words. That is the whole product. It is not a video editor, it does not clone the voice of a named real person, and it does not place your avatar into footage. What it gives you is the piece most short ads are missing: a human face saying a specific sentence to the camera. ## Which avatar model should you pick? Three models, and the only real trade is resolution against credits per second. All three are capped at 15 seconds of speech per generation, with a 3-second minimum. | Model | Resolution | Credits per second | 15-second clip | Cost on the $49.99 plan | Cost on the $199.99 plan | |---|---|---|---|---|---| | InfiniteTalk | 480p | 1 | 15 credits | $1.07 | $1.00 | | Kling Avatar | 720p | 2 | 30 credits | $2.14 | $2.00 | | Kling Avatar Pro | 1080p | 4 | 60 credits | $4.28 | $4.00 | Pick this way: - **InfiniteTalk at 480p** is for auditions. Run the script here first at 1 credit a second, watch it back, and find out whether the wording sounds like a person before you pay four times as much. - **Kling Avatar at 720p** is the working default. It is sharp enough for a phone screen, which is where every one of these ads is watched. - **Kling Avatar Pro at 1080p** is for the clip you are actually going to spend money distributing, or for anything that gets cropped, zoomed or cut into a longer edit. Auditioning at 480p and re-running the keeper at 720p costs 45 credits instead of 30, which is about $3.21 on the $49.99 plan. That is cheaper than shipping a 1080p clip whose second sentence does not scan. ## How long can the script be? Fifteen seconds of speech is a hard ceiling on one generation, and that works out at about 35 words. Write past it and you are writing a second clip whether you meant to or not. | Clip length | Words at conversational pace | InfiniteTalk 480p | Kling Avatar 720p | Kling Avatar Pro 1080p | |---|---|---|---|---| | 3 seconds, the minimum | about 7 | 3 credits | 6 credits | 12 credits | | 5 seconds | about 12 | 5 | 10 | 20 | | 8 seconds | about 18 | 8 | 16 | 32 | | 12 seconds | about 28 | 12 | 24 | 48 | | 15 seconds, the maximum | about 35 | 15 | 30 | 60 | Two things follow from that table. **You are charged on the real audio, not on your estimate.** A script you counted as 12 seconds that voices at 14 costs 28 credits on Kling Avatar 720p, not 24. Counting words is the cheapest way to keep the bill predictable. **A 30-second ad is two clips.** Generate the first 15 seconds, generate the second 15 seconds, and join them. Write it as two halves with a clean break between them, not as one paragraph cut in the middle, because the seam lands wherever your sentence ends. The hook plus demo mode does a related job: a talking hook of up to 8 seconds joined to a demo clip in one 1080 x 1920 file, for 2 credits. ## How do you write 15 seconds that holds? Four rules, then three scripts you can use as they are. - **Front-load the problem.** The first three seconds decide the rest. Never open with the brand name. - **Short sentences sync better.** A long unbroken clause makes the mouth movement look mechanical. Full stops give the model somewhere to breathe. - **One claim, not three.** Thirty-five words holds one argument. Pick the strongest. - **Say what happens next.** A clip with no ask is a clip that gets watched and forgotten. ### 1. Language learning app, 35 words - **Hook, 0 to 3 seconds:** "Three hundred day streak and you still cannot order lunch." - **Body, 3 to 12 seconds:** "This drills speaking from day one. Ten-minute calls with an AI tutor that corrects pronunciation and sends the words you missed." - **Call to action, 12 to 15 seconds:** "Lesson link is below." ### 2. Home insurance comparison, 35 words - **Hook:** "Your home insurance renewed itself again." - **Body:** "This site reads your current policy, matches the cover, and shows what six insurers charge for the same thing. Four minutes, no phone calls." - **Call to action:** "Compare yours from the link." ### 3. Spreadsheet course, 34 words - **Hook:** "Half the jobs you want ask for advanced spreadsheets." - **Body:** "This is twelve hours of lessons with real data files, pivot tables, lookups and dashboards. You finish with three interview projects." - **Call to action:** "Course link is below." More angles by category sit in the [UGC ads index](https://www.makeviral.ai/ugc-ads), and [SaaS UGC ads](https://www.makeviral.ai/ugc-ads/saas) and [mobile app UGC ads](https://www.makeviral.ai/ugc-ads/mobile-apps) both cover the spokesperson shape in detail. ## When does a talking avatar beat a voiceover video? A talking avatar and a faceless voiceover video solve different problems, and the cost gap between them is wide enough to matter. | | Talking avatar | [AI voiceover video](https://www.makeviral.ai/tools/ai-voiceover-video-generator) | |---|---|---| | On screen | A face, lip-synced to the words | Gameplay footage or generated visuals | | Maximum length | 15 seconds per clip | 90 seconds | | Cost | 30 credits for 15 seconds at 720p | 10 credits per video | | Strongest at | The opening line, an objection, a direct ask | Explaining, teaching, stacking an argument | | Weakest at | Anything needing more than 35 words | Making a viewer feel spoken to | Use the avatar when the message is short and personal: a spokesperson video for a landing page, the first three seconds of an ad, an answer to the one objection that kills your checkout. Use the faceless route when the product needs a minute of context, and the whole [faceless UGC generator](https://www.makeviral.ai/tools/ugc-video-generator) wing is built for that. The two combine well. A talking hook of 8 seconds costs 16 credits on Kling Avatar 720p. Join it to an animated product clip for 2 more, and you have a face opening an ad that a faceless format finishes. That is the cheapest use of an avatar there is, and the [AI UGC ad generator page](https://www.makeviral.ai/tools/ai-ugc-ad-generator) has the full hook test built around it. ## What makes lip sync look wrong? Lip sync quality depends on the portrait more than on the model. A bad portrait at 1080p looks worse than a good portrait at 480p, because the resolution just shows you the problem in more detail. **Portraits that sync cleanly** - The face is front-facing, or close to it, and roughly centered. - The light is even. No hard shadow cutting across the mouth. - Nothing covers the lower face: no hand on the chin, no mug, no microphone, no scarf. - The mouth is closed or slightly open. A wide fixed grin gives the model very little to work with. - The head is at a normal distance. An extreme close-up or a tiny face in a wide room both sync worse. **Portraits that do not** - A profile or a three-quarter turn past about 30 degrees. - Heavy motion blur, low resolution, or a heavily filtered face. - Two people in frame. The model needs one face to animate. If a generated portrait syncs badly, regenerate the portrait rather than the video. The portrait costs 2 credits on Nano Banana 2 or 1 credit on Z-Image. A 15-second video costs 30. Fixing the cheap end first is always the right order, and the [pricing page](https://www.makeviral.ai/pricing) has the full credit table if you want to plan a batch. ## What an AI spokesperson may not claim The avatar is not a person. It never used your product, never held it before you uploaded the image, and has no experience to describe. Keep the script on the product and this is simple. - **Pitch, do not testify.** "Twelve hours of lessons and three interview projects" is a product fact. "This got me my job" is a claim by somebody who does not exist. - **Claims stay inside what the product does.** Features, specifications, price, what is included. No invented outcomes, no star ratings on screen, no numbers you cannot source. - **No borrowed authority.** Do not dress the avatar as a doctor, a lawyer or an official, and do not write a line that implies a qualification. - **Check the current disclosure rules for the platform you are buying on before you run it.** They change, and they differ by country. Verified on 17 September 2026: - The FTC's [Disclosures 101 for Social Media Influencers](https://www.ftc.gov/business-guidance/resources/disclosures-101-social-media-influencers) says plainly: "You can't talk about your experience with a product you haven't tried." It also says that in a video "the disclosure should be in the video and not just in the description uploaded with the video". - The FTC's [final rule banning fake reviews and testimonials](https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials), announced August 2024, covers testimonials that "misrepresent that they are by someone who does not exist, such as AI-generated fake reviews". This is not legal advice. Read the rules for your platform and your country before the clip goes anywhere. ## What this generator does not do - **No clip longer than 15 seconds.** The ceiling is on the generation, not on the finished ad, so longer ads are stitched from parts. - **No 16:9 and no long form.** Output is vertical, 1080 x 1920. This is a short-form tool. - **No API.** Everything runs in the browser, so a hundred unattended renders is not something you can script. - **No ad account integration, no scheduling, no posting.** You download the file and upload it to Meta Ads Manager, TikTok Ads Manager or wherever else yourself. - **No voice cloning of a named real person.** Each creator carries its own voice. - **No analytics.** Numbers live in the platform you posted to. If any of those are load-bearing for you, find out now rather than after the first month. If none of them are, the loop is short: cast a creator once, then every new clip is a new 35 words. The [AI influencer generator hub](https://www.makeviral.ai/ai-influencer-generator) lists the rest of the wing, [how to make UGC ads with AI](https://www.makeviral.ai/blog/how-to-make-ugc-ads-with-ai) is the full workflow, [how to go viral on TikTok](https://www.makeviral.ai/blog/how-to-go-viral-on-tiktok) covers the organic side, and the [Arcads alternatives page](https://www.makeviral.ai/alternatives/arcads-ai) is the comparison if you are still shortlisting. ## AI Avatar Video Generator examples Real outputs from the app, exactly as they download. - **Character explainer, why we sleep** (Explainer mode, the preset character Nova, Kling Avatar, script written from the topic, Ink captions). Brief: "Why do humans need sleep?". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/explainer-why-sleep.mp4, 17 s. 38 credits. Made 2026-09-18. - **UGC ad, portable blender** (Hook + demo, a preset creator, Kling Avatar hook, animated product shot, Hormozi captions). Brief: "Okay, I did not expect this tiny thing to crush ice.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/ugc-blender.mp4, 11 s. 18 credits. Made 2026-09-18. - **UGC ad, LED face mask** (Hook + demo, a preset creator, Kling Avatar hook, product shot animated with Seedance 2.0 Mini, Submagic captions). Brief: "Three weeks in and my skin has never been this calm.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/ugc-led-mask.mp4, 11 s. 28 credits. Made 2026-09-18. - **UGC ad, sunset lamp** (Hook + demo, a preset creator, Kling Avatar hook, animated product shot, Hormozi captions). Brief: "Nine dollars, and my room looks like this now.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/ugc-sunset-lamp.mp4, 10 s. 16 credits. Made 2026-09-18. - **Product photo, blender in hand** (Product in hand, the creator photo plus the product photo, 1K, 9:16). Brief: "She holds the blender bottle upright in one hand at chest height in her kitchen.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/ugc-blender-still.png. 2 credits. Made 2026-09-18. - **Product photo, LED mask in hand** (Product in hand, the creator photo plus the product photo, 1K, 9:16). Brief: "She holds the white face mask flat in one hand at chest height in the bathroom. Her own face stays fully visible above the mask.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/ugc-led-mask-still.png. 2 credits. Made 2026-09-18. - **Product photo, lamp in hand** (Product in hand, the creator photo plus the product photo, 1K, 9:16). Brief: "She holds the small black lamp in one hand in her apartment and its warm orange light falls on the wall behind her.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/ugc-sunset-lamp-still.png. 2 credits. Made 2026-09-18. - **Maya, a made creator** (Nano Banana 2, 1K, 9:16, one prompt and no photo of a real person). Brief: "Maya, 26, curly dark hair, warm smile, cream knit sweater, selfie at arm's length in a cozy sunlit apartment.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/influencer-persona-maya.png. 2 credits. Made 2026-09-17. - **The creator holds the product** (Nano Banana 2, 1K, the creator photo plus one product photo). Brief: "Upload the product photo. The app writes the prompt, keeps the face and keeps the label readable.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/influencer-product-in-hand.png. 2 credits. Made 2026-09-18. - **The creator says the hook** (Kling Avatar, 720p, AI voice, lip synced to the script). Brief: "Okay, I have to tell you about this cleanser. Two weeks in and my skin is actually calm. Go look at it.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/influencer-talking-hook.mp4, 7 s. 16 credits. Made 2026-09-17. - **The product shot, animated** (Veo 3.1 Lite, 720p, 8 s, the product photo becomes the first frame). Brief: "She shows the product to the camera, slight smile, subtle hand motion, natural indoor light, static phone camera.". File: https://makeviralai.s3.us-east-1.amazonaws.com/examples/influencer/influencer-demo-animate.mp4, 8 s. 6 credits. Made 2026-09-18. ## What the ai avatar video generator does for you - **Lip sync from a portrait.** A voice reads your script, then the 9:16 portrait is animated to match that exact audio track. - **Three models, three price points.** InfiniteTalk at 480p for 1 credit a second, Kling Avatar at 720p for 2, Kling Avatar Pro at 1080p for 4. - **Scripts written for you.** Draft from the product name, description and angle, then cut it back to 35 words before generating. - **3 to 15 seconds per clip.** Billing follows the real audio length, so a 14-second read costs 28 credits on Kling Avatar 720p. - **Hook plus demo stitch.** Join a talking hook of up to 8 seconds to a demo clip in one 1080 x 1920 file for 2 credits. ## How it works 1. **Cast a creator.** A preset, a written description or an uploaded face photo. The portrait is 2 credits, or 1 on Z-Image. 2. **Write 35 words.** Hook, one claim, one ask. Short sentences sync better than long unbroken clauses. 3. **Pick a model and generate.** Audition at 480p for 1 credit a second, then re-run the keeper at 720p or 1080p. ## Who it is for SaaS teams · Mobile app marketers · Course creators · Direct-to-consumer brands · Agencies · Insurance and finance · Coaches and consultants · Affiliate marketers ## Frequently asked questions ### What is an AI avatar video generator? It is a tool that animates a portrait to match an audio track, so a still face appears to speak. In MakeViral you cast a creator, write a script of 35 words or fewer, and the creator's voice reads it while the 9:16 portrait is lip-synced to that recording. ### How long can a talking avatar video be? From 3 to 15 seconds of speech per generation. Fifteen seconds is a hard ceiling, which is about 35 words at conversational pace. A 30-second video is two clips generated separately and joined, so write it as two halves with a clean break rather than one long paragraph. ### Which avatar model should I use? InfiniteTalk at 480p costs 1 credit per second and is best for auditioning a script. Kling Avatar at 720p costs 2 credits per second and is the sensible default for phone viewing. Kling Avatar Pro at 1080p costs 4 credits per second, for clips you will distribute or crop. ### What does a 15-second talking avatar video cost? Thirty credits on Kling Avatar at 720p, since it bills 2 credits per second of speech. On the $49.99 plan for 700 credits that is about $2.14. On the $199.99 plan for 3,000 credits it is about $2.00. The same clip is 15 credits at 480p and 60 credits at 1080p. ### Why does my lip sync look stiff? Almost always the portrait. Front-facing, evenly lit faces with nothing covering the mouth sync cleanly. Profiles, hard shadows, a hand on the chin, heavy filters or two people in frame do not. Regenerate the portrait for 2 credits before you regenerate a 30-credit video. ### Can I use this for a spokesperson video on my website? Yes, within the limits. The output is vertical 1080 x 1920 and at most 15 seconds per clip, so it suits a short explainer block or a social ad rather than a minute-long corporate film. Longer pieces have to be stitched from several clips. ### Can the avatar say it used my product? No. The person does not exist and has no experience with anything, so keep the script to what the product does. The FTC's influencer guidance, checked in September 2026, says you cannot talk about experience with a product you have not tried. Check your platform's disclosure rules too. ### Does MakeViral post the video for me? No. There is no posting, no scheduling, no ad account integration and no API. You download the 1080 x 1920 MP4 and upload it yourself. There is also no analytics here, so performance numbers stay in the platform where you ran the clip. ## Viral AI video templates - [Hotel Lobby AI: Make the Hotel Lobby Trend Video](https://www.makeviral.ai/templates/ai-hotel-lobby-trend): You and a friend, or your two pets, trade a rap into one hanging mic in a bright orange booth. Step into the real clip or the viral cat version with the original song, or get a new 10-second version with an original beat. - [AI Dog Dancing Video: Make Your Pet Dance](https://www.makeviral.ai/templates/ai-dancing-dog-video): Your dog or cat stands up on its hind legs at home and does a happy, bouncy dance. One pet photo in, a 10-second clip with an original beat out. - [80s AI Trend: Your 1985 Photo and Home Video](https://www.makeviral.ai/templates/80s-ai-trend): One selfie becomes a faded 1985 snapshot with big hair and a date stamp. Stop there for the photo, or animate it into a 10-second camcorder home video with sound. - [All AI video templates](https://www.makeviral.ai/templates) ## Related tools and guides - [AI UGC Creator](https://www.makeviral.ai/tools/ai-ugc-creator): Pick a preset, describe a person in words, or upload a face photo. The portrait, the name and the voice save together, so one creator can front a whole campaign. - [AI UGC Ad Generator](https://www.makeviral.ai/tools/ai-ugc-ad-generator): Cast a creator, upload your product, write 35 words. You get a vertical ad with a person on camera in minutes, not in a two-week creator cycle. - [TikTok Voice Generator for AI Voiceover Videos](https://www.makeviral.ai/tools/ai-voiceover-video-generator): The narrator-over-gameplay format: a written script becomes a voiced, captioned vertical video with a background clip, in one pass. - [How to Make UGC Ads With AI (2026 Workflow)](https://www.makeviral.ai/blog/how-to-make-ugc-ads-with-ai): Pick the angle, cast an AI creator, write 15 seconds, render the clip, add the demo, test three hooks. See the workflow and build your first ad. - [How to Go Viral on TikTok: What Actually Moves Reach](https://www.makeviral.ai/blog/how-to-go-viral-on-tiktok): What TikTok says about recommendations, 15 hooks that hold the first two seconds, a retention checklist and a 14-day plan. Start your first video. - [Every ai influencer generator](https://www.makeviral.ai/ai-influencer-generator) ## A talking avatar in the time it takes to write 35 words Kling Avatar at 720p costs 2 credits per second of speech, so a 15-second clip is 30 credits, about $2.14 on the $49.99 plan. Cancel anytime, 14-day refund window on unused credits. [Make a talking video](https://www.makeviral.ai/influencer)