A commercial is not a product clip with music on top. It needs shots that cut together, light that sells the product, a voice, words on screen, and an end card, all inside 15 seconds. Most AI commercial generator pages show you a template gallery. We made the same spot for a made-up sparkling water four different ways and wrote down what each one cost, how long it took, and what it got wrong.
If you need launch spots, seasonal variants, or ten versions of one ad for testing, this is the short version: a one-click generator gives you a finished ad in about four minutes, a script-to-ad model follows your voiceover word for word, and building it shot by shot was the cheapest at $1.35 for 15 seconds and the only route where we controlled every frame. Every clip below ran on Modellix with one API key, so the only thing that changed between routes was the model.
AI Commercial Generator Options at a Glance
| Route | Start with | You give it | You control | 15 s spot |
|---|---|---|---|---|
| One click |  Vidu One-Click AD-Film |
1 to 7 product photos | Length, ratio, a one-line brief | $3.00 |
| Script to ad | Vidu Q3 AD |
A short script + product and scene images | Scenes, voiceover lines | $1.80 |
| One call with a voice |  Seedance 2.5 R2V |
Reference images + a shot list | Shots, spoken line, sound | $3.47 |
| Shot by shot |  Kling V3 I2V +  Gemini TTS |
One keyframe per shot | Every frame, word, and cut | $1.35 |
Pick the route by how much you need to control. One click is fastest, but you cannot choose the shots. Shot by shot takes longer to set up, and you approve every frame.
Make Your First AI Commercial on One API Key
Every model in this test runs from the same account: one-click ads, script-to-ad, voice, and single shots. No subscription; you pay for the seconds you generate.
Get API KeyWe Made the Same 15-Second Commercial Four Ways
The product is SOLVA Yuzu Sparkling Water, a brand we invented so no real label gets mangled. One packshot from Nano Banana 2, then four relit versions of the same can from Nano Banana 2 Edit: $0.46 for all five images. Every route started from these images and the same brief: open on the can in the dark, reveal it with light, cut to a sunny picnic and a splash, end on the line "Bright on the inside."
Route 1: One click, four photos in
Vidu One-Click AD-Film took four of the images and a one-line brief and returned a 15-second, six-shot ad with music and a voiceover it wrote itself: "Sulva yuzu sparkling water, crisp, bright, sun-kissed refreshment, bright on the inside." It kept our end line and the label in every shot. It also cut to a picnic blanket covered in grapefruit, which is the most appetizing fruit in the test, just not the one on the can.
Route 2: Script to ad
Vidu Q3 AD takes a screenplay plus up to 14 reference images tagged as characters, scenes, or props. We gave it the can as a prop, three scene images, and this script:
Scene 1 INT. BLACK STUDIO - NIGHT A thin line of light sweeps across @SolvaCan in the dark. Droplets sparkle. NARRATOR (V.O.): Some days need a little light. Scene 2 EXT. PICNIC TABLE - GOLDEN HOUR @SolvaCan stands on the table in warm sunlight beside an ice bucket. NARRATOR (V.O.): Real yuzu. Real bubbles. Scene 3 INT. SPLASH SET - DAY Sparkling water splashes around @SolvaCan against a cyan background. NARRATOR (V.O.): SOLVA. Bright on the inside.
It spoke all three lines word for word and planned six shots, including a macro of droplets on the yuzu icon we never asked for. Two things to plan around. There is no duration setting, so our three short scenes became a 28-second ad, and at $0.12 per second, length is cost. And in the splash shot, the small "355 ml" turned into made-up characters, while the big SOLVA type stayed perfect. It was also the slowest route at 15.5 minutes.
Route 3: One call with a voice
Seedance 2.5 R2V takes reference images and a shot list, and with audio on it adds the voice and sound in the same call. We asked for two shots in 8 seconds: a light sweep in the dark, then the picnic table. Both shots landed and the label stayed sharp. The narrator's "yuzu" came out mushy enough that our transcript read it as "YouTube", so listen before you ship a product name.
Route 4: Shot by shot, cut ourselves
We animated each lit image with Kling V3 I2V (5 seconds, 720p, sound off, $0.34 a shot), voiced the 15-word script with Gemini 3.1 Flash TTS for $0.01, and cut 15 seconds with captions and an end card in a short ffmpeg script. No model chose a shot for us, so nothing surprised us either.
Side by side
| Route | What came back | Label | Time to result | Cost |
|---|---|---|---|---|
| One-Click AD-Film |
15 s, 6 shots, music + its own voiceover | Kept | 256 s | $3.00 |
| Vidu Q3 AD |
28 s, 6 shots, our 3 lines word for word | Fine print lost | 15.5 min | $3.38 list |
| Seedance 2.5 R2V |
8 s, 2 shots, voiceover | Kept | 152 s | $1.85 |
| Shot by shot | 15 s, 4 shots, TTS voice, captions, end card | Kept | 62 to 83 s a shot | $1.35 |
- Big type survives everywhere. SOLVA and "Yuzu Sparkling Water" stayed readable in every route. Small print is the first thing to go.
- The automatic routes add things. Grapefruit, a droplet macro, 13 extra seconds. Fine for a draft, worth a frame-by-frame check before it airs.
- Shot by shot costs the least and takes the most work. You write the cuts, but every frame is one you picked.
Product Ad Lighting: 4 Setups, Same Can
Light is what makes a product shot look like an ad and not a listing photo. We set the light on the still image first, then animated it, because an image-to-video model starts from the first frame you give it. Four setups, same can, same Kling V3 settings:
Same can, four lighting setups, each animated once. Generated on Modellix: Kling V3 I2V, 5 seconds, 720p, sound off.
Lighting prompts that held
Describe the light in the image edit, then describe a motion that shows it off in the video prompt. The label sentence goes last in both.
| Setup | Use it for | Light words that held | Motion that suits it |
|---|---|---|---|
| Softbox | E-commerce hero, end card | "large softbox key light from the left, white fill card on the right, soft even shadow" | Slow orbit, 20 degrees or less |
| Rim light | Premium, launch teaser, night | "pure black background, two strip lights behind, thin bright rim highlights" | A band of light sweeping across the label |
| Golden hour | Lifestyle, summer, outdoors | "low warm sun behind and to the right, gentle lens flare" | Drifting leaf shadows, slow handheld push-in |
| Hard light | Energy, sports, drinks | "hard direct light from the upper right, crisp shadows, saturated background" | A splash that builds, from a dry first frame |
The video prompt pattern we used for all four:
[Light setup] on [the product]. [One small motion that shows the light: a sweep, an orbit, drifting shadows]. [Camera move]. Keep the [shape] and the '[LABEL TEXT]' label text exactly as shown.
Three lighting shots that went wrong
- A frozen splash plays out. Our splash still showed water mid-air, so the model let it fall: by the end of the clip it was ripples. Start from a dry can and prompt the splash, or use the splash image as the end frame on a model that accepts one, such as Veo 3.1 I2V.
- An orbit shows what you never designed. The slow orbit turned the softbox can far enough that Kling invented back-of-can fine print and a barcode. Keep orbits small, or give the model a back-label image.
- The sweep came late. The rim light sweep only crossed the label near the end of the clip. Ask for it "in the first half", or trim the clip so the sweep lands on your voiceover beat.
How to Make a Commercial With AI, Step by Step
This is the shot-by-shot route, the one that gave us the most control for the least money.
- Write the voiceover first. Our 15-word script took 8.3 seconds to say at an unhurried pace, which left room for pauses and the end card. Keep a 15-second spot under about 25 words.
- Make one keyframe per shot from one packshot. Use an image edit model to put the same product in each scene and set the light there. Same product in, same product out.
- Animate each shot for 5 seconds at 720p. One small motion per shot, sound off, label sentence last. Our four shots took 62 to 83 seconds each.
- Voice it. Text to speech with a short style note ("warm, confident, unhurried"); ours cost under one cent. MiniMax Speech 2.8 HD is the option if you need finer control over pitch and emotion.
- Cut it. Trim each shot to its beat, add captions and an end card, then a licensed music track. We did it in ffmpeg; the Modellix plugin's FFmpeg and MoviePy skills can do the same edit from a coding agent.
For more shot types to put in the edit (close-ups, dolly moves, multi-shot scenes), see the prompt recipes in our AI movie maker guide. If the ad needs a person talking to camera, the AI UGC video generator guide covers that format.
Run it without code
Every model above has a playground on its Modellix model page: inputs on the left, the finished clip on the right. Open Vidu One-Click AD-Film, add up to seven product photos, write one line about the product, and press Run; the price per second shows before you spend anything. For shot-by-shot work, do the same on Kling V3 I2V with one keyframe at a time.
What an AI Commercial Costs
Pay-per-second pricing makes a commercial easy to budget: seconds times rate, plus retries. Here is each route scaled to a 15-second spot, and to ten variants for an ad test. The Q3 AD and Seedance figures use the per-second rate from our runs; the shot-by-shot figure is what our four shots and voiceover actually cost.
| Route | 15 s spot | 10 variants for an A/B test |
|---|---|---|
| Shot by shot (Kling V3 + TTS) |
$1.35 | $13.50 |
| Vidu Q3 AD |
$1.80 | $18.00 |
| Vidu One-Click AD-Film |
$3.00 | $29.96 |
| Seedance 2.5 R2V, with audio |
$3.47 | $34.65 |
Retries are part of the price. We ran every route once. Plan on one or two extra takes per shot in production, which still keeps the shot-by-shot route under $5 for a 15-second spot.
For single product shots rather than full ads, our AI product video generator test compares six image-to-video models on one product photo, with the cost per usable clip.
Why Modellix is the best API for AI commercials
One API key, every route in this test
- Mix routes in one project. A one-click draft for the pitch, Q3 AD for the scripted version, Kling V3 shots for the final cut, all from one script and one bill.
- No vendor accounts. Vidu, ByteDance, Kling, and Google models through one key, with the same submit, poll, and download flow.
- Costs you can audit. Every model page lists its per-second price and every task reports what it billed. That is how this guide knows what each spot cost.
Six ways to run these models
Playground
No code. Sign in, open a model page such as Vidu One-Click AD-Film, upload your image or paste a prompt, and the clip, time, and cost show up in your browser. Best for picking a shortlist.
Coding agent, no install
Paste one sentence into Claude Code, Codex, or Cursor and let it read the docs:
My Modellix API key is mdlx-••••••••••••••••••••••••. Read https://docs.modellix.ai/get-started/index.md, https://docs.modellix.ai/kling/kling-v3-i2v.md and https://docs.modellix.ai/google/gemini-3-1-flash-tts.md. For each keyframe in ./shots, make a 5-second 720p clip with kling/kling-v3-i2v (audio off) using the prompt in the matching .txt file, voice script.txt with google/gemini-3.1-flash-tts, then cut everything into a 15-second 16:9 commercial with ffmpeg. Run one task at a time and report the cost of each.
Swap in your own key from the console, and only share it with agents you trust.
Agent Skill
Install the skill once and your agent reads each model's schema and fills in the request fields itself. Name the model in your prompt; if you don't, it asks which one to run. Works in any Agent Skills host. Setup is in the Skill guide.
npx skills add https://github.com/Modellix/modellix-plugin --skill modellix-design
Plugin
If your host has a plugin marketplace (Claude Code, Codex, Cursor), the plugin adds six skills (generation, image prompts, social cards, infographics, and local edits with FFmpeg or MoviePy) plus the Media Generation MCP, so commands like /modellix:video and /modellix:social-card work out of the box. In Claude Code, run /plugin marketplace add Modellix/modellix-plugin, then /plugin install modellix@modellix. Other hosts are in the Plugin guide.
CLI
For scripts and cron jobs. One command submits the task, waits, and prints the result URL; modellix-cli model batch takes a JSONL file for a whole list. See the CLI guide.
npm install --global modellix-cli
modellix-cli model run --model-slug vidu/one-click-ad-film --body-file request.json --wait
REST API
Submit, keep the task_id, poll, then download before the result URL expires. Field names vary by provider, so check each model's schema in the API guide.
Start from a prompt that already works
Prompts from earlier Modellix tests that already work for ads and brand shots. Generated on Modellix.
Macro product hero
“A minimalist unbranded steel wristwatch with a plain white dial rotates on a black glass pedestal while water droplets fall in slow motion, crisp specular highlights, macro commercial lighting. No logos.”
Spotlight sweep
“Slow dolly-in on a woman sitting alone at a dim jazz bar counter, warm tungsten light. As a spotlight sweeps across her face, her expression shifts from calm to quiet surprise. Shallow depth of field, 35mm film look.”
Multi-shot story in one call
“Shot 1: close-up of a chef's hands plating a dessert. Shot 2: wide shot of the busy restaurant kitchen. Shot 3: the chef looks up and smiles at the camera. Warm cinematic lighting.”
One-click skincare ad
“A premium 15-second ad for ORVEL Daily Glow Serum: calm morning skincare routine, clean product hero shots, keep the label text exact.”
FAQ About AI Commercial Generators
What is the best AI commercial generator?
For a finished ad from product photos in a few minutes, Vidu One-Click AD-Film. For a scripted voiceover read word for word, Vidu Q3 AD. For full control at the lowest cost, Kling V3 I2V shots plus text to speech, cut yourself. If you want to try all of them without signing up four times, Modellix is the fastest and most cost-effective way in: one account, one key, pay per second.
Can I make a commercial using AI?
Yes. In our test, every route produced a usable 8 to 28 second ad from one generated product image, and the product name stayed readable. The work that is left is checking small print and adding licensed music.
How much does a 15-second AI commercial cost?
On Modellix, our shot-by-shot spot cost $1.35, a one-click ad cost $3.00, and a script-to-ad spot comes to about $1.80 at $0.12 per second. Budget for retries on top.
Is there a free AI commercial generator?
Free tiers on template tools usually add a watermark or cap the resolution. Modellix has no subscription; you pay for the seconds you generate, and our ads came back without watermarks.
Which AI is free for commercial use?
"Free" and "cleared for commercial use" are separate questions. Check the output terms of the model provider and the ad platform's rules on AI-generated content, and keep your prompts and source images so you can show how each ad was made.
How long does it take to make an AI commercial?
One generation took 1 to 15.5 minutes in our test, from a single Kling shot to the 28-second Q3 AD. Writing the script and checking frames took longer than any model.
Are some commercials AI-generated?
Yes. Coca-Cola's AI-made holiday ads are the best-known example, and they drew as much criticism as attention. Most AI in advertising is less visible: drafts, variants, and product shots that go through the same review as any other footage.
Related articles
AI Music Video Generator [Tested]: Beyond the Singing Head
A 10-second song, five AI models, and a 20-second music video that isn't a singer holding a mic. Every prompt, every cost, and what broke.
AI Product Video Generator [Tested]: 1 Photo, 6 Models
See real AI product video clips, the cost per usable clip, and the prompt that stops a model from reshaping your product, before you spend a cent.
Storyboard AI [Tested]: Shot List to Animatic for Under $2
We ran storyboard AI end to end: one shot list through 3 image models and 5 video models. Free shot list template, prompts, results, and real costs.
ComfyUI Image to Video [Tested]: Local Wan vs No-GPU API
We ran ComfyUI image to video two ways: local Wan 2.2 on a 16 GB Mac (it failed) and an API node with Kling and Veo. Setup, workflow JSON, and costs.