AI Model Prices Checked Against the Bill: Video, Image and Voice at No Markup
What Kling, Veo, MiniMax, Nano Banana and ElevenLabs list per second, per image and per 1,000 characters, a worked example of each, and what the render really cost us. Where they match, and the two rows that do not.
Made with Hermoso
We compared what each AI model maker lists with what a real render cost us, line by line. On video, image and voice, 13 of the 15 models we could check came out at the list price, or within a fraction of a cent once the prompt is counted. Two Grok models cost more than xAI's list, and they're marked below. Everything else in this post is the working.
This is what "no markup on models" rests on. You can check any row yourself, because makers publish these numbers.
How to read the tables
- List price is the maker's own number, with its unit, read on November 1, 2026 unless the row says otherwise.
- Standard ask is one small, normal job: a short clip at 720p, one image at about 1K, or one 116-character voiceover line.
- Cost us is the provider cost on that render's row in our own usage ledger, not an estimate. Each is one real render on the date shown.
Video: priced per second of output
You pay for the seconds you get back, so a 5-second clip costs five times the rate. Resolution changes the rate, which is why the tables name it.
| Model | List price | Standard ask | At list | Cost us | Source |
|---|---|---|---|---|---|
| Kling 4.0 | $0.157 a second, 720p, no input video | 5 seconds | $0.785 | $0.785 (Nov 1) | Kling's API price table (read Oct 29) |
| MiniMax H3 Max | $0.08 a second, 768P | 5 seconds | $0.40 | $0.40 (Nov 1) | MiniMax pay-as-you-go page |
| Veo 3.1 | $0.40 a second, with sound | 6 seconds | $2.40 | $2.40 (Nov 1) | Google (read Oct 20) |
| Veo 3.1 Fast | $0.10 a second, with sound | 4 seconds | $0.40 | $0.40 (Oct 20) | Google (read Oct 20) |
| Veo 3.1 Lite | $0.05 a second, with sound | 4 seconds | $0.20 | $0.20 (Oct 22) | Google (read Oct 20) |
| Gemini Omni 1.1 | About $0.10 a second at 720p, as Google works it out from tokens | 4 seconds | About $0.40 | $0.41 (Oct 20) | Gemini API pricing page |
| Grok Imagine 1.5 Lite | $0.02 a second, no resolution stated | 6 seconds | $0.12 | $0.18 (Nov 1) | xAI's model page |
Update, 8 November 2026: Google released Gemini Omni 1.2 on 5 November and it replaced Omni 1.1 on Hermoso. It lists at $12 per million video tokens, about $0.07 a second at 720p (about $0.28 for the same 4 seconds), and Google routes Omni 1.1 requests to 1.2 from 3 December 2026. Our first Omni 1.2 render, 4 seconds at 720p, cost us $0.28 (Nov 7).
Seedance 2.5 is missing because we couldn't read a public list price to hold it against. A 6-second clip at 720p cost us $1.40, or about 23 cents a second.
Images: priced per image, sometimes plus the prompt
Google prices the picture and also counts the words you send, so a real render lands a little over the headline number. That's the difference on the Google rows here.
| Model | List price | Standard ask | At list | Cost us | Source |
|---|---|---|---|---|---|
| Nano Banana 2 Lite | $0.0336 per 1K image | 1 image | $0.0336 | $0.0336 (Nov 1) | Gemini API pricing page |
| Nano Banana 2.1 | $0.0336 per 1K image | 1 image | $0.0336 | $0.037 (Oct 31) | Gemini API pricing page |
| Nano Banana Pro | $0.134 per 1K or 2K image | 1 image | $0.134 | $0.1375 (Oct 20) | Gemini API pricing page |
| Grok Imagine 2.0 | $0.04 an image | 1 image | $0.04 | $0.06 (Nov 1) | xAI's model list |
Voice: priced per 1,000 characters
A line of 116 characters is about seven seconds of speech. A 30-second ad script runs about 450 characters, so multiply by four.
| Model | List price | 116 characters at list | Cost us | A 450-character script | Source |
|---|---|---|---|---|---|
| Eleven v4 | $0.08 per 1,000 | $0.00928 | $0.00928 (Oct 20) | 3.6 cents | ElevenLabs API pricing |
| Speech 2.8 HD | $0.10 per 1,000 | $0.0116 | $0.0116 (Oct 20) | 4.5 cents | MiniMax pay-as-you-go page |
| Speech 2.8 Turbo | $0.06 per 1,000 | $0.00696 | $0.00696 (Nov 1) | 2.7 cents | MiniMax pay-as-you-go page |
| Grok Voice | $15 per million | $0.00174 | $0.00174 (Oct 20) | 0.7 cents | xAI's model list |
The two rows that miss
Grok Imagine 1.5 Lite and Grok Imagine 2.0 both run about 50% above xAI's list price.
What you pay on Hermoso
On Starter, Pro and Studio the model part of a render is charged at the model cost, in steps of 2 cents. So a 3.4-cent image is 4 cents, a $0.785 clip is 80 cents, and a very short voice line rounds up to 2 cents. The exact price is on the button before anything runs. Plans are on the pricing page.
For comparison, a lot of multi-model tools sell credits that work out at about double the model price. AI video generator pricing, explained shows how to work out any tool's real rate in five steps, and no markup on models covers why we dropped ours.
Price vs quality, model by model
Price only tells you half of it, so we also ran the same ad on each model and show every take:
- 10 AI video models, one prompt, and
- 11 image models for ads, and
- 6 voices, one ad script, plus the four below
Four more voices, same script
These four were not in our first voice test. Same line, one take each, on November 1, 2026. Where Whisper, running locally, heard something other than the script, the caption says so.
Hermoso is an AI marketing agent and AI ad generator with 100+ image, video, voice and writing models. Ask for an ad in plain words, pick the model or let the agent pick, and the app shows the exact price before anything runs. Start free at app.hermoso.ai.
Questions people ask
How much does Kling 4.0 cost?
Kling lists $0.157 a second at 720p with no input video, so a 5-second clip is $0.785. Our render on November 1, 2026 cost exactly that.
How much does a Veo 3.1 clip cost?
Google listed $0.40 a second with sound when we read it in October 2026, so 6 seconds is $2.40, which is what a render cost us on November 1, 2026. Veo 3.1 Lite is $0.05 a second.
How much does ElevenLabs cost per 1,000 characters?
Its API lists $0.08 per 1,000 characters for Eleven v4, so a 450-character, 30-second script is about 3.6 cents.
Does Hermoso charge more than the model maker?
No markup on models: on Starter, Pro and Studio the model part of a render is charged at the model cost, in steps of 2 cents. Two Grok models cost us more than xAI's own list, and this post marks them.
Why do Google image renders cost slightly more than the list price?
Google prices the output image and also counts the input tokens in your prompt, so a real render lands a little above the per-image figure.