How Much Does AI Video Cost? Prices for 20 Seconds, 1 Minute and 5 Minutes, With and Without Voice

A 20-second AI video costs from about $0.40 to $11 to generate in November 2026, a one-minute video from about $1.20 to $34, and a five-minute video from about $6 to $171. That is the price of one take through each vendor’s API. The spread depends on the model, the resolution and whether the model also generates sound. A voiceover is the cheap part: five minutes of synthetic narration costs between $0.07 and $0.50.
These numbers are our own calculation from the price pages of Google, xAI, Kling, Runway, Luma, MiniMax, ByteDance and Alibaba, all checked on November 15, 2026. Below are the full tables for 20 seconds, 1 minute and 5 minutes, the cost of sound and voice, what a monthly subscription buys, and image prices for the same services. One finding surprised us: the same model can cost 2.4 times more depending on where you buy it.
AI video costs at a glance

| Question | Short answer |
|---|---|
| How is AI video priced? | Mostly per second of finished video through an API, or in credits through a monthly plan |
| What does a 20-second video cost? | About $0.40 (Grok Imagine Video 1.5 Lite at 480p) to $11.38 (Seedance 2.5 at 1080p), one take |
| What does a 1-minute video cost? | About $1.20 to $34. Veo 3.1 Fast at 1080p with sound costs $7.20 |
| What does a 5-minute video cost? | About $6 to $171. Veo 3.1 Fast at 1080p with sound costs $36 |
| How much does a voiceover add? | $0.07 to $0.50 for five minutes of synthetic speech through an API |
| How much does native sound add? | It depends on the model: 20% more for Veo 3.1 Fast, 50% more for Kling 3.0, double for Veo 3.1 |
| Is a subscription cheaper than the API? | Sometimes. Google’s own plan is cheaper than its API for Veo 3.1, but Runway’s Standard plan costs 2.4 times its API price |
| Can I still buy Sora? | No. OpenAI shut down the Sora API on September 24, 2026, with “no one-to-one replacement” |
| What does an AI image cost? | From under 1 cent to about 24 cents each through an API, depending on model, quality and size |
How AI video is priced
Three things decide the bill, and all of them are easy to miss on a pricing page.
You pay per second of output. Almost every video API now charges for each second of finished video. Google lists Veo 3.1 on Vertex AI at $0.40 per second at 720p or 1080p with sound, so an 8-second clip costs $3.20. Some services sell credits instead, which you then convert into seconds. Runway’s API, for example, charges 12 credits a second for Gen-4.5, and a credit costs $0.01.
A long video is many short clips. No model generates five minutes in one go. Veo 3.1 makes 8-second clips, Gemini Omni 1.2 Flash makes 3 to 15 seconds, Grok Imagine Video up to 15 seconds, and Luma’s Ray3.2 up to 10 seconds. A five-minute video therefore takes 20 to 40 separate generations, which then have to be edited together.
You pay for every take you keep or throw away. Kling, ByteDance and Alibaba say they do not charge for generations that fail or are blocked by a safety filter. But a clip that renders fine and simply looks wrong is billed in full. If you keep one take in three, the real cost is three times the numbers in our tables. That ratio is our assumption for illustration, not a vendor figure.
There is also a difference between native sound and a voiceover. Some models, such as Veo 3.1, Gemini Omni and Kling 3.0, can generate sound together with the picture: ambient noise, effects and short lines of dialogue. A narration over a whole video is usually made separately with a text-to-speech model and laid over the edited clips. They are priced very differently, as you will see below.
What changed in 2026: Sora is gone, Veo moved
If you searched for “Sora pricing” or “Veo 3 pricing”, the answer changed this year.
- OpenAI Sora. OpenAI shut down the Videos API with
sora-2andsora-2-proon September 24, 2026, and its deprecations page lists “No one-to-one replacement”. The last listed price forsora-2was $0.10 per second at 720p. The Sora app closed earlier, in April 2026, according to OpenAI’s help center as quoted in search results (the help center itself blocked our requests). - Google Veo. Veo 3.1 is no longer available in the Gemini API: Google shut down the preview models there on October 22, 2026, and points developers to Gemini Omni Flash or to Veo 3.1 on Google Cloud’s Vertex AI, which Google now calls Gemini Enterprise Agent Platform. In the Flow app, Veo 3.1 Fast and Quality retire on December 5, 2026.
Veo 3 pricing now
Veo 3.1 is still sold through Google Cloud’s Vertex AI. These are the per-second prices from Google’s pricing page (Google writes “/1 count”, and Runway resells Veo 3.1 at the same $0.40 a second):
| Model | Video only | Video with sound | 4K with sound |
|---|---|---|---|
| Veo 3.1 | $0.20 per second | $0.40 per second | $0.60 per second |
| Veo 3.1 Fast | $0.08 (720p), $0.10 (1080p) | $0.10 (720p), $0.12 (1080p) | $0.30 per second |
| Veo 3.1 Lite | $0.03 (720p), $0.05 (1080p) | $0.05 (720p), $0.08 (1080p) | not offered |
| Gemini Omni 1.2 Flash (Gemini API) | not offered | about $0.07 (720p), $0.104 (1080p) | about $0.209 per second |
Google says Gemini Omni 1.2 Flash costs “approximately $0.07 per second” at 720p, or about $0.70 for a 10-second clip, with native audio included.
Our calculation: 20 seconds, 1 minute and 5 minutes
We multiplied each published per-second price by 20, 60 and 300 seconds. These are API prices for one take of every shot, without re-rolls, editing or tax.
| Service | Model | Resolution | Price per second | 20 seconds | 1 minute | 5 minutes |
|---|---|---|---|---|---|---|
| xAI | Grok Imagine Video 1.5 Lite | 480p | $0.02 | $0.40 | $1.20 | $6.00 |
| Google (Vertex AI) | Veo 3.1 Lite, no sound | 720p | $0.03 | $0.60 | $1.80 | $9.00 |
| Google (Vertex AI) | Veo 3.1 Lite, with native sound | 720p | $0.05 | $1.00 | $3.00 | $15.00 |
| Luma | Ray3.2 | 720p | $0.06 | $1.20 | $3.60 | $18.00 |
| Google (Gemini API) | Gemini Omni 1.2 Flash (sound included) | 720p | $0.07 | $1.40 | $4.20 | $21.00 |
| MiniMax | MiniMax-H3 | 768p | $0.08 | $1.60 | $4.80 | $24.00 |
| Google (Vertex AI) | Veo 3.1 Fast, no sound | 1080p | $0.10 | $2.00 | $6.00 | $30.00 |
| Google (Gemini API) | Gemini Omni 1.2 Flash (sound included) | 1080p | $0.104 | $2.08 | $6.24 | $31.20 |
| Kling | Kling 3.0, no sound | 1080p | $0.112 | $2.24 | $6.72 | $33.60 |
| Google (Vertex AI) | Veo 3.1 Fast, with native sound | 1080p | $0.12 | $2.40 | $7.20 | $36.00 |
| Runway | Gen-4.5 | not stated | $0.12 | $2.40 | $7.20 | $36.00 |
| Alibaba | Wan 3.0 (30% off promotion) | 1080p | $0.14 | $2.80 | $8.40 | $42.00 |
| xAI | Grok Imagine Video 1.5 | 720p | $0.14 | $2.80 | $8.40 | $42.00 |
| ByteDance (BytePlus) | Seedance 2.0 | 720p | $0.15 | $3.00 | $9.00 | $45.00 |
| Kling | Kling 3.0, with native sound | 1080p | $0.168 | $3.36 | $10.08 | $50.40 |
| Google (Vertex AI) | Veo 3.1, no sound | 1080p | $0.20 | $4.00 | $12.00 | $60.00 |
| Luma | Ray3.2 | 1080p | $0.24 | $4.80 | $14.40 | $72.00 |
| Google (Vertex AI) | Veo 3.1, with native sound | 1080p | $0.40 | $8.00 | $24.00 | $120 |
| ByteDance (BytePlus) | Seedance 2.5 | 1080p | $0.569 | $11.38 | $34.14 | $171 |

The chart shows the same five minutes at one take per shot. The cheapest and the most expensive option differ 28 times, and resolution matters as much as the brand: Luma’s Ray3.2 costs four times more at 1080p than at 720p.
Some notes on the sources. Kling’s prices are for Kling 3.0 from Kling’s own API price list, which also has a cheaper Kling 3.0 Turbo ($0.14 a second at 1080p with sound); Kling 4.0 is sold through resellers such as Runway and Pika. The cheapest row, Grok Imagine Video 1.5 Lite, is priced at 480p; at 720p it costs $0.03 a second, the same as Veo 3.1 Lite without sound. Seedance prices come from ByteDance’s BytePlus price examples for 5-second clips. Alibaba lists Wan 3.0 at $0.20 a second for 1080p with a limited-time 30% discount, which gives the $0.14 in the table. Luma bills in 5-second blocks. Runway does not publish a resolution for Gen-4.5.
A 20-second video
Twenty seconds is two or three generations: three 8-second clips with Veo 3.1, or two clips with Gemini Omni, Grok or Luma. Through an API this costs:
- Under $1: Grok Imagine Video 1.5 Lite at 480p ($0.40), Veo 3.1 Lite without sound ($0.60).
- $1 to $3: Veo 3.1 Lite with sound ($1), Gemini Omni 1.2 Flash at 720p with sound ($1.40), Veo 3.1 Fast at 1080p with sound ($2.40), Kling 3.0 at 1080p without sound ($2.24).
- $8 and more: Veo 3.1 with sound ($8), Seedance 2.5 at 1080p ($11.38).
A 20-second voiceover adds about 3 cents or less with any of the voice models in the table further down.
A 1-minute video
One minute is four clips of 15 seconds with Gemini Omni, or eight clips with Veo 3.1. A typical mid-range choice, Veo 3.1 Fast at 1080p with sound, costs $7.20. Gemini Omni 1.2 Flash at 720p with sound costs $4.20, and the premium Veo 3.1 with sound costs $24. With a narration from ElevenLabs’ v3 or v4 model the total grows by about 8 cents.
A 5-minute video
Five minutes is where the choice of model starts to matter in real money: $6 with Grok Imagine Video 1.5 Lite at 480p, $21 with Gemini Omni at 720p, $36 with Veo 3.1 Fast or Runway Gen-4.5, $120 with Veo 3.1 with sound and $171 with Seedance 2.5 at 1080p. It also means 20 to 40 separate clips, so the time spent on prompts, re-rolls and editing usually matters more than the bill. If you keep one take in three, a five-minute Veo 3.1 Fast video with sound costs about $108 rather than $36.
For a five-minute video where a person talks to the camera, a generative video model is often the wrong tool. Avatar services are built for that and price by the minute; see the section on presenter videos below.
With sound or without: what voice and audio cost
“With sound” can mean two different things, and the price difference between them is large.
Native sound: from 20% to double
Native sound is generated by the video model together with the picture. On Vertex AI it is a separate price line, and the markup varies:
| Model | Without sound | With sound | Markup |
|---|---|---|---|
| Veo 3.1 Fast, 1080p | $0.10 per second | $0.12 per second | +20% |
| Kling 3.0, 1080p | $0.112 per second | $0.168 per second | +50% |
| Veo 3.1 Lite, 720p | $0.03 per second | $0.05 per second | +67% |
| Veo 3.1, 1080p | $0.20 per second | $0.40 per second | +100% |
For Veo 3.1, sound adds $60 to a five-minute video. Gemini Omni 1.2 Flash has no silent option: its price already includes audio.
A voiceover: cents, not dollars
A narration made with a text-to-speech model costs almost nothing next to the video. We counted 1,000 characters of text per minute of speech, which matches ElevenLabs’ own equation of 10,000 credits with “~10 minutes” of audio, and about 150 words a minute.
| Service | Voice model | 20 seconds | 1 minute | 5 minutes |
|---|---|---|---|---|
| Gemini 3.8 Flash TTS (price until Dec 31, 2026) | $0.005 | $0.013 | $0.068 | |
| xAI | Text to Speech | $0.005 | $0.015 | $0.075 |
| Google Cloud | Chirp 3: HD voices | $0.010 | $0.030 | $0.15 |
| ElevenLabs API | Flash / Turbo models | $0.013 | $0.040 | $0.20 |
| ElevenLabs API | Eleven v3 / v4 | $0.027 | $0.080 | $0.40 |
| MiniMax | speech-2.8-hd | $0.033 | $0.10 | $0.50 |
Sources: ElevenLabs API pricing ($0.04 to $0.08 per 1,000 characters), the Gemini API price list ($9 per million audio tokens at 25 tokens a second, rising to $18 on January 1, 2027), Google Cloud Text-to-Speech ($30 per million characters for Chirp 3 HD), xAI ($15 per million characters) and MiniMax ($100 per million characters).
If you prefer a subscription, the ElevenLabs Starter plan costs $6 a month for about 30 minutes of speech. OpenAI’s older voice models, tts-1, tts-1-hd and gpt-4o-mini-tts, are deprecated and shut down on January 6, 2027.
So which is cheaper for a narrated video? In most cases, silent video plus a separate voiceover. A five-minute Veo 3.1 video at 1080p costs $60 silent plus $0.40 for an ElevenLabs narration, against $120 with native sound. Native sound is worth paying for when you need effects and dialogue that match the picture, not a voice over it.
Presenter and avatar videos
For a five-minute video with a talking presenter, avatar services charge by the minute and include the voice:
| Service | Plan | Price per month | What it includes |
|---|---|---|---|
| Synthesia | Starter | $19 (or $14 a month billed yearly; $9 for the first month) | Up to about 12 minutes a month on top of the free 10 minutes |
| Synthesia | Pro | $89 (or $59 a month billed yearly; $49 for the first month) | Up to about 60 minutes a month on top of the free 10 minutes |
| HeyGen | Creator | $29 (or $24 a month billed yearly) | 600 credits: about 19 minutes of Avatar IV video or about 150 minutes of Avatar III |
| HeyGen | Pro | $49 | 1,000 credits, up to 4K |
Sources: Synthesia pricing, HeyGen pricing and HeyGen’s help article on credits (Avatar IV uses 31 credits a minute with the video look, Avatar III uses 4). On these numbers a five-minute presenter video costs roughly $4 to $6 of a Synthesia plan and fits several times into a HeyGen Creator plan.
Our calculation: the same model, different sellers
Many video models are sold in several places: by the developer’s own API, by resellers such as Runway and Pika, and inside monthly plans. We compared what 60 seconds of the same model costs in each, assuming a subscriber uses every monthly credit on that model.

| Model | Where you buy it | Price per second | 1 minute |
|---|---|---|---|
| Runway Gen-4.5 | Runway API (12 credits a second at $0.01) | $0.12 | $7.20 |
| Runway Gen-4.5 | Runway Standard plan ($15 for 625 credits) | $0.29 | $17.28 |
| Kling 4.0, 1080p | Runway API, list price (23.5 credits a second) | $0.235 | $14.10 |
| Kling 4.0, 1080p | Pika Starter plan ($10 for 600 credits, 118 credits per 5 seconds) | $0.39 | $23.60 |
| Kling 4.0, 1080p | Runway Standard plan (118 credits per 5 seconds) | $0.57 | $33.98 |
| Veo 3.1 with sound | Google AI Pro plan in Flow ($19.99 for 1,000 credits, 100 credits per 8 seconds) | $0.25 | $14.99 |
| Veo 3.1 with sound | Vertex AI or Runway API | $0.40 | $24.00 |
| Gemini Omni 1.2 Flash, 720p | Google AI Pro plan in Flow (3 credits a second) | $0.06 | $3.60 |
| Gemini Omni 1.2 Flash, 720p | Gemini API | $0.07 | $4.20 |
Two conclusions. First, Runway’s Standard plan costs 2.4 times its own API for both Gen-4.5 and Kling 4.0, so a heavy user is better off with the API or a bigger plan. Second, subscriptions are not always the expensive option: Google’s own plan undercuts its API for Veo 3.1 and Gemini Omni. Note that Veo 3.1 Quality leaves Flow on December 5, 2026, and that until November 30 Runway sells Kling 4.0 at a promotional 18.8 credits a second through its API and at 25% off in its plans.
What a monthly subscription buys
If you do not want to deal with an API, here is roughly how much video the entry plans give each month, again assuming you spend every credit on one model:
| Plan | Price per month | Model | Video per month | Price per second |
|---|---|---|---|---|
| Google AI Pro | $19.99 | Gemini Omni 1.2 Flash, 720p, with sound | About 5.5 minutes (333 seconds) | $0.06 |
| Luma Plus | $30 | Ray3.2, 720p, 5-second clips | About 8 minutes (500 seconds) | $0.06 |
| Luma Plus | $30 | Ray3.2, 1080p, 5-second clips | About 2 minutes (125 seconds) | $0.24 |
| Pika Starter | $10 | Pika 2.5, 720p | About 2.5 minutes (150 seconds) | $0.067 |
| Adobe Firefly Standard | $9.99 | Firefly video | “Up to” 100 seconds (20 five-second videos) | $0.10 |
| Runway Standard | $15 | Gen-4.5 | About 52 seconds | $0.29 |
Sources: Google AI plans and Flow credits, Luma, Pika, Adobe Firefly, Runway. Google Flow also gives everyone 50 free credits a day. Unused credits do not roll over at Google Flow, so a half-used plan doubles your real price per second.
Kling sells app plans from $6.99 to $127.99 a month as first-subscription offers, according to Kling’s own blog, but it publishes credit costs per second only for 4K, so we could not convert them. The plans page loads its prices with a script we could not read.
Midjourney prices
Midjourney sells only subscriptions, with no public API. These are the plans from Midjourney’s documentation:
| Plan | Monthly | Billed yearly | Fast GPU time |
|---|---|---|---|
| Basic | $10 | $8 a month | 3.3 hours |
| Standard | $30 | $24 a month | 15 hours |
| Pro | $60 | $48 a month | 30 hours |
| Mega | $120 | $96 a month | 60 hours |
Midjourney makes video too: one job gives four 5-second silent clips. An HD video job uses about 26 minutes of fast GPU time, which on the Standard plan works out to about 4 cents per second of output. Midjourney videos have no sound.
How much AI images cost
Images are much cheaper than video. These are API prices per image, as published:
| Service | Model | Price per image | 100 images |
|---|---|---|---|
| OpenAI | gpt-image-2, 1024 x 1024: low / medium / high quality | $0.006 / $0.053 / $0.211 | $0.60 / $5.30 / $21.10 |
| Nano Banana 2.1, 1K | $0.0336 | $3.36 | |
| Nano Banana 2, 1K | $0.067 | $6.70 | |
| Nano Banana Pro, 1K or 2K | $0.134 | $13.40 | |
| Google (Vertex AI) | Imagen 4 Fast / Imagen 4 / Imagen 4 Ultra | $0.02 / $0.04 / $0.06 | $2 / $4 / $6 |
| Black Forest Labs | FLUX 3, 1K | $0.048 | $4.80 |
| xAI | Grok Imagine Image / Image 2.0 (1K, low quality) | $0.02 / $0.04 | $2 / $4 |
| ByteDance | Seedream 5.5, up to about 1.5K | $0.02 | $2 |
| Kling | Kling Image 3.0 | $0.028 | $2.80 |
| Recraft | V4.1, raster | $0.035 | $3.50 |
| Stability AI | Stable Image Core / Ultra | $0.03 / $0.08 | $3 / $8 |
Sources: OpenAI image generation guide, Gemini API pricing, Vertex AI pricing, Black Forest Labs, xAI, BytePlus, Kling, Recraft, Stability AI.
OpenAI’s newest image models, gpt-image-2.6-sunburst and gpt-image-2.6-flare, are priced by tokens ($30 per million output tokens). OpenAI has no per-image price table for them, only a calculator, so we listed gpt-image-2, which has one. Note that gpt-image-2 is deprecated and shuts down on February 10, 2027. Ideogram’s API price table did not load for us.
On subscriptions, Adobe Firefly Standard ($9.99) says it covers “up to 400 images” a month, and Midjourney’s Standard plan and above include unlimited image generations in the slower Relax mode.
How to pay less for AI video
- Draft cheap, finish expensive. Test prompts on a Lite or Fast model at 720p, then render the final takes on the better model. Veo 3.1 Lite costs $0.03 a second, Veo 3.1 with sound $0.40.
- Add the voice separately. For narration, a silent video plus a text-to-speech voiceover is far cheaper than native sound: $0.40 against $60 extra for five minutes on Veo 3.1.
- Pick resolution last. Moving from 720p to 1080p quadruples Luma’s price and more than doubles Seedance 2.5’s.
- Check who sells it. The same model can cost 2.4 times more on one platform than on another, so compare the developer’s API, resellers and plans.
- Use up subscriptions or skip them. Credits that do not roll over raise your real price per second. For irregular use, pay as you go is usually cheaper.
- Watch promotions and deadlines. Wan 3.0 is on a 30% discount, Runway’s Kling 4.0 promotion ends November 30, Veo 3.1 Fast and Quality leave Flow on December 5, and Gemini’s voice price doubles on January 1, 2027.
For the cost of text models, see our guide to reducing LLM API costs and the API pricing table. Chatbot subscriptions are compared on our AI chatbot pricing page.
Frequently asked questions
How much does it cost to make a 1-minute AI video?
Through an API, from about $1.20 (Grok Imagine Video 1.5 Lite at 480p) to about $34 (Seedance 2.5 at 1080p) for one take of every shot. A common mid-range choice, Veo 3.1 Fast at 1080p with native sound, costs $7.20 a minute. Re-rolls multiply this.
How much does a 5-minute AI video cost?
From about $6 to $171 in generation fees, by our calculation from November 2026 price pages. Veo 3.1 Fast at 1080p with sound costs $36, Gemini Omni 1.2 Flash at 720p $21. Five minutes takes 20 to 40 separate clips, so editing time is a bigger cost than the bill for most people.
How much does Veo 3 cost?
Veo 3.1 on Google’s Vertex AI costs $0.40 a second with sound and $0.20 without at 720p or 1080p, $0.10 to $0.12 for Veo 3.1 Fast with sound and $0.05 for Veo 3.1 Lite with sound at 720p. Veo is no longer in the Gemini API, where Google now offers Gemini Omni 1.2 Flash at about $0.07 a second.
How much does Sora cost?
Sora is no longer for sale. OpenAI shut down the Sora API on September 24, 2026, and lists no direct replacement. Its last listed price was $0.10 a second at 720p.
Is it cheaper to add a voiceover or use native sound?
A voiceover is much cheaper. Five minutes of text-to-speech costs $0.07 to $0.50, while native sound on Veo 3.1 adds $60 to the same five minutes. Native sound makes sense for effects and dialogue that must match the picture.
Do failed AI video generations cost money?
Kling, ByteDance’s Seedance and Alibaba’s Wan say they do not charge for failed or blocked generations. A clip that renders but does not look right is billed in full.
What is the cheapest AI video generator?
By API price per second in our table, Grok Imagine Video 1.5 Lite at $0.02 a second at 480p. At 720p it costs $0.03, the same as Veo 3.1 Lite without sound. Cheaper models usually mean lower resolution and less control, so test them on your own prompts first.
How much does an AI image cost?
Through an API, from about half a cent (OpenAI gpt-image-2 at low quality) to about 21 cents (gpt-image-2 at high quality) or 24 cents (Nano Banana Pro at 4K). Most popular models cost 2 to 7 cents an image.