Every AI video model we run, with specs and credit costs
12 models from 8 vendors, in one account and on one credit balance. This page lists what each model can actually do (maximum clip length, the resolutions we price, how it handles sound, whether it needs a reference image) and exactly what a render costs in credits.
We publish this because the model you pick is the biggest single lever on what a render costs, and that choice is impossible to make from vendor marketing pages. The numbers here come straight out of the same pricing table the app charges against, so what you read below is what your balance is debited.
The spread is real: the cheapest priced render in the catalogue is Grok Video at 5 seconds, 480p, 50 credits, and the most expensive is Seedance 2.0 at 15 seconds, 4K, 1,985 credits. Picking the right model for the job is the single biggest lever on what your ad testing costs.
Model comparison table
Listed alphabetically by model name, because no other order is neutral. Credits per render is the range across every clip length and resolution we price for that model: the full per-combination table is on each model page.
| Model | Vendor | Max clip | Resolutions priced | Audio | Reference image | Free trial | Credits per render |
|---|---|---|---|---|---|---|---|
| Grok Video | xAI | 10s | 480p · 720p | Generated with the picture | Required | Included | 50 to 140 |
| Happy Horse 1.1 | Alibaba | 15s | 720p · 1080p | Generated with the picture | Required | Included | 140 to 550 |
| Kling 2.6 | Kuaishou | 10s | 720p · 1080p | Not stated | Required | Included | 125 to 260 |
| Kling 3.0 | Kuaishou | 15s | 720p · 1080p · 4K | Generated with the picture | Required | Paid plans only | 210 to 1,170 |
| OmniHuman 1.5 | ByteDance | 30s | 720p · 1080p | From a voice track we generate | Required | Paid plans only | 490 to 980 |
| Pruna Avatar | Pruna | 15s | 1080p | Spoken by the model | Required | Included | 92 to 368 |
| Seedance 1.5 | ByteDance | 12s | 1080p | Generated with the picture | Required | Included | 100 to 245 |
| Seedance 2.0 | ByteDance | 15s | 720p · 1080p · 4K | Generated with the picture | Required | Paid plans only | 140 to 1,985 |
| Sora 2Vendor end date announced | OpenAI | 12s | 1080p | Generated with the picture | Not required | Paid plans only | 80 to 245 |
| Sora 2 ProVendor end date announced | OpenAI | 12s | 720p · 1080p | Generated with the picture | Not required | Paid plans only | 245 to 1,225 |
| VEED Fabric 1.0 | VEED | 120s | 720p | From a voice track we generate | Required | Paid plans only | 460 to 920 |
| Veo 3.1 Fast | 8s | 720p · 1080p | Generated with the picture | Required | Included | 120 to 245 |
How to read this
- Max clip is the longest single render the model will make. The ceiling runs from 8 seconds up to 120 seconds (VEED Fabric 1.0), so a script that will not fit inside one clip narrows the shortlist fast. On OmniHuman 1.5, Pruna Avatar and VEED Fabric 1.0 the length is not a setting: the clip runs as long as the speech, and the credit cost follows it. Where you do pick the length, the ceiling is 15 seconds (Happy Horse 1.1, Kling 3.0 and Seedance 2.0).
- Resolutions priced is what we actually sell, not what a vendor page advertises. Only Kling 3.0 and Seedance 2.0 carry a 4K tier. On Kling 2.6 and OmniHuman 1.5 the two tiers cost the same and reach the model as the same request, so the resolution there is a label rather than a setting.
- Audio splits three ways. Grok Video, Happy Horse 1.1, Kling 3.0, Pruna Avatar, Seedance 1.5, Seedance 2.0, Sora 2, Sora 2 Pro and Veo 3.1 Fast make sound in the same pass as the picture. OmniHuman 1.5 and VEED Fabric 1.0 render onto a voice track instead, which we generate from your script first, so the input you supply is still an image plus a script. Kling 2.6 reads "Not stated": the behaviour is not confirmed on our side, and a guess there would be worse than a gap.
- Reference image matters when you have nothing to work from. Sora 2 and Sora 2 Pro will generate from a written prompt alone; everything else expects a still.
- Free trial is a real restriction, not a soft one. The app refuses a trial render on Kling 3.0, OmniHuman 1.5, Seedance 2.0, Sora 2, Sora 2 Pro and VEED Fabric 1.0, and it refuses any trial render costing more than 250 credits, whichever model it is on.
- Credits per render is a range because length and resolution both move the price. The cheapest entry point in the catalogue is Grok Video at 50 credits.
What credits cost
One balance, every model. A plan grants a monthly pool of credits and each render subtracts its own cost from that pool, so a month of cheap iteration and a month of 4K finals draw on the same allowance at very different rates. That is the honest way to price a catalogue where one render can cost 50 credits and another can cost 1,985.
| Plan | Monthly price | Credits included |
|---|---|---|
| Starter | $49 per month | 5,000 credits |
| Growth | $99 per month | 12,000 credits |
| Agency | $199 per month | 25,000 credits |
Annual billing is 30% off the same plans. Full details, including what happens when a balance runs out, are on the pricing page. Every plan starts free for 3 days.
What the free 3-day trial covers
Not every model, and we would rather say so here than at the Generate button. 6 of the 12 run during the trial: Grok Video, Happy Horse 1.1, Kling 2.6, Pruna Avatar, Seedance 1.5 and Veo 3.1 Fast. Trial renders are also capped at 250 credits each, which puts the longer renders on Happy Horse 1.1, Kling 2.6 and Pruna Avatar out of reach until a plan starts.
Kling 3.0, OmniHuman 1.5, Seedance 2.0, Sora 2, Sora 2 Pro and VEED Fabric 1.0 run on any paid plan instead. The account, the library and the credit balance are the same on both sides, so moving to a plan unlocks them without rebuilding anything you set up during the trial.
Each model page states its own answer, and the table above has a column for it.
Head to head
Picking between two specific models? Each comparison puts the two credit tables on the same page. See all model comparisons.
All 12 models
Frequently asked questions
How many AI video models can I use on UGC Vids AI?
12, from 8 different vendors (xAI, Alibaba, Kuaishou, ByteDance, Pruna, OpenAI, VEED and Google). They all draw on one credit balance inside the same account, so switching model costs you a setting rather than a second subscription.
Which AI video model is the cheapest?
Grok Video has the lowest starting cost of any model here: its cheapest priced render is 5 seconds, 480p, 50 credits. Cost scales with clip length and resolution, so the cheapest model for a 15 second 4K master is not the cheapest model for a 5 second hook.
Which models render in 4K?
Kling 3.0 and Seedance 2.0. Every other model we run tops out at 1080p or below. On all 2 of them the 4K tier costs more credits than the same clip length at 1080p, so it is worth reserving for finals rather than tests.
Which model makes the longest single clip?
VEED Fabric 1.0 has the highest ceiling at 120 seconds, though you do not set that length directly: the clip runs as long as the speech does. Where you choose the clip length yourself, Happy Horse 1.1, Kling 3.0 and Seedance 2.0 run to 15 seconds. Longer ads are built by generating several clips and joining them, which is why clip ceilings matter more than they look.
Which models can I use on the free 3-day trial?
Grok Video, Happy Horse 1.1, Kling 2.6, Pruna Avatar, Seedance 1.5 and Veo 3.1 Fast. Trial renders are also capped at 250 credits each, which puts the longer renders on Happy Horse 1.1, Kling 2.6 and Pruna Avatar out of reach until a plan starts. Kling 3.0, OmniHuman 1.5, Seedance 2.0, Sora 2, Sora 2 Pro and VEED Fabric 1.0 run on any paid plan rather than on the trial. The account and the credit balance are the same either way, so nothing has to be rebuilt when you switch.
Do I need a reference image to generate a video?
For most models, yes: a reference image is part of the input. Sora 2 and Sora 2 Pro are the exceptions and will generate from a written prompt alone.
Which models generate their own audio?
Grok Video, Happy Horse 1.1, Kling 3.0, Pruna Avatar, Seedance 1.5, Seedance 2.0, Sora 2, Sora 2 Pro and Veo 3.1 Fast produce speech and picture in the same pass, with no separate voiceover step. OmniHuman 1.5 and VEED Fabric 1.0 work the other way round: they render onto a voice track, which we generate from your script before the render starts, so from your side the input is still an image plus a script. We publish no audio answer for Kling 2.6: the behaviour is not confirmed on our side, and a gap is better than a claim we cannot stand behind.
Do I pay per model or per plan?
Per plan. Starter is $49 a month with 5,000 credits, Growth is $99 a month with 12,000 credits, Agency is $199 a month with 25,000 credits. Every model draws from the same balance, so the cost of a render depends on the model, the clip length and the resolution you pick, not on which plan you are on.
Related reading: what a finished AI video ad costs across every model, the best model for lip sync, and our full comparison library. Prices and plans live on the pricing page.