Use cases · Video ads

The AI Video Ad Generator You Can Run by API

Revana puts Veo, Kling, Seedance, Wan and the voice and music models behind one API key and one request shape, so cinematic spots, story ads and product cutaways come from your own tools or an AI agent. You pay per second of video from prepaid credit, and a render that fails costs nothing: test ten hooks, keep the two that win.

Ad styles you can make on Revana

Cinematic spots with sound

Veo 3.1 renders dialogue, ambience and music with the picture in one pass, from $0.20 a second at 720p and $0.40 a second with sound (prices as of November 2026). Kling 4.0 makes clips of up to 30 seconds with sound from $0.135 a second at 720p.

Story ads with a consistent cast

Seedance 2.5 takes reference images and makes clips of 4 to 30 seconds with sound, from $0.09 a second. Pass pictures of your product and cast to keep them consistent from shot to shot; up to 30 fit in one request. It suits a hook, a problem and a payoff.

Talking creators

A presenter speaking to camera works well for direct-response ads, and Revana has the avatar, voice and lip-sync models for it. The full creator workflow is on the UGC video page; use only faces and voices of people who agreed to appear.

Product hero shots

Turn a product photo into a short hero shot for a carousel or a retargeting ad on the same key, on a low-cost or a premium model with the same request. The catalogue workflow is on the product video page.

Test more hooks without burning the budget

  1. Write ten hooks. Ask a language model on the same key, such as Claude or GPT on the OpenAI-compatible chat API, for ten opening lines and shot lists. Tokens bill from the same balance as your video.
  2. Draft them cheaply. Render first drafts on Wan 3.0 from $0.045 a second at 480p, or Gemini Omni 1.2 Flash from $0.024 a second at 360p. Both add sound, so you can judge the full ad.
  3. Render the winners. Send the winning prompts to Veo 3.1, Kling 4.0 or Seedance 2.5 at 1080p. On Revana that is a one-word change to the model field; prompts, references and pipeline stay the same.
One test round, priced on Revana
StepKlipModelEachTotal
Drafts10Wan 3.0 480p, 5 s$0.23$2.30
Finals2Kling 4.0 1080p, 10 s$2.00$4.00
Round12$6.30

Prices as of November 2026. Quote any request free at /v1/generations/quote before you run it.

Charged once, for what arrives

Revana holds the quoted price when you submit and releases it if the render fails. A failed job costs $0.00, so failures never show up on your bill and the cost per finished clip stays the number you quoted.

Another route takes over

Most video models on Revana have more than one route. If one errors, times out or is full, the next takes over without any retry code from you, so a bulk run keeps moving through a provider outage.

Idempotency and a price cap

Send an Idempotency-Key with each job so a resend after a timeout never makes a second video, and set max_cost_usd so no single render can cost more than you planned.

Restage an ad format that already works

Your people, products and places

Send Nemesis a clip you own or licensed plus up to 8 pictures, and it plays the same moves, timing and camera with your people, products and places. Nemesis (aka Genjutsu) costs $0.17 a second at 720p, so a 10 second recast is $1.70; Nemesis Lite does clips of up to 15 seconds for $0.095 a second.

Rights

Recast only footage you filmed, licensed or have permission to change, and only cast people who agreed to it. Keep the rights for each source clip on file; ad platforms can ask for them.

Music, voiceover and sound on the same bill

Native sound

Veo 3.1, Kling 4.0, Seedance 2.5 and Wan 3.0 can render sound with the picture. Describe the voice lines and ambience in the prompt, or leave sound off where it costs extra and you plan to add your own track.

Music beds

Suno V6 makes a full song or an instrumental from an idea, your lyrics or a style, and returns two tracks per generation for $0.08. Ask for a 15 or 30 second bed per campaign.

Voiceover in many languages

ElevenLabs V3, V4, Turbo 2.5 and Multilingual V2 run from $0.21 per 1,000 characters. Localise one ad for several markets by changing the script and the voice; the picture can stay the same.

Vertical, square and widescreen from one prompt

Seedance 2.5, Kling 4.0 and Veo 3.1 all render 9:16; set the aspect ratio in the request and the model frames the shot for a phone screen instead of you cropping a widescreen clip. Seedance 2.5 and Kling 4.0 also render 1:1 for feed placements and marketplaces, and 16:9 for YouTube and your own site.

Satu request
# One Idempotency-Key per job, made once: resend with the same $KEY after a timeout, a 429 or a 5xx
KEY=$(uuidgen)
curl https://api.revana.si/v1/generations \
  -H "Authorization: Bearer $REVANA_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $KEY" \
  -d '{
  "model": "kling-4.0",
  "input": {
    "prompt": "Hook: a hand slams an alarm clock, cut to the product on a sunlit desk, upbeat ambience",
    "duration": 10,
    "resolution": "1080p",
    "aspect_ratio": "9:16"
  },
  "max_cost_usd": 2
}'

Make ads from Claude or ChatGPT

From Claude or ChatGPT

Connect the Revana MCP server to Claude, ChatGPT, Cursor or VS Code and ask for three vertical hooks for your product. The agent picks the models, quotes the price and waits for your yes, then hands back download links when the clips are ready.

One balance for the whole team

A strategist can make test clips in a chat while a developer builds the pipeline on the same Revana account. Both draw on one prepaid balance, and every job shows its price before it runs, so no one spends credit by surprise.

A ready-made ad skill

The drama ad skill makes a 25-second ad shot like a premium streaming show, with Seedance 2.5 and Nano Banana Pro, from Claude Code and your Revana key.

FAQ

What is an AI video ad generator API?

It is a way to make video ads from code, an automation or an AI agent instead of an editor. On Revana one key and one request shape reach Veo, Kling, Seedance, Wan and the voice and music models, billed per second from prepaid credit.

How much does an AI video ad cost?

It depends on the model, resolution and length. Ten 5 second drafts on Wan 3.0 at 480p cost $2.30, and two 10 second finals on Kling 4.0 at 1080p cost $4.00 on Revana (prices as of November 2026). Failed renders cost $0.

Can the ads have sound?

Yes. Veo 3.1, Kling 4.0, Seedance 2.5 and Wan 3.0 can render sound with the picture; describe the voice lines and ambience in the prompt. Suno V6 adds music and ElevenLabs a voiceover on the same key.

Which aspect ratios work for ads?

Seedance 2.5, Kling 4.0 and Veo 3.1 all render 9:16 natively, and Seedance 2.5 and Kling 4.0 also render 1:1 and 16:9. Set the ratio in the request rather than cropping a widescreen clip, so faces and products stay in frame.

Can I copy a competitor's ad?

No. Restage only footage you filmed, licensed or have permission to change, and cast only people who agreed to it. Revana is not for impersonating real people or copying someone else's ad.

Do I need to code?

For a large test matrix, yes, or an automation tool that sends HTTP requests. For a few ads, the Revana MCP server lets Claude or ChatGPT write the hooks, quote the clips and make them in a chat.

Can I run the ads commercially?

That depends on each model maker's terms and on the rights you hold for the people, products and footage in the ad. Check both before you spend on media.

Make your first ad variants today

Create your Revana key, send ten hooks as drafts and render the two that win. Running ads at volume? The Revana team can set up volume pricing and invoicing on the same key.