Creatify Boreal is a new AI video model that charges one cent for every second of video it produces. Creatify Labs, the research arm of the ad-video company Creatify, launched it on 15 September 2026. It is the company’s second in-house model after Aurora, its talking-avatar model, and it turns a text prompt or a still image into a short clip with sound, rendering a five-second clip in about five seconds.
“A video generation API that renders in realtime at a groundbreaking rate of 1¢ per second,” is how the launch release sums it up. Creatify says Boreal rivals top-tier closed models on the measures that matter for advertising, at a fraction of the price and about 40 times the speed. It is built on LTX-2.5, the open video model from Lightricks, rather than trained from scratch.
Below we explain what Creatify Boreal is, then check its price, speed and benchmark claims against Creatify’s own pages, where several figures do not agree with each other. We also cover how to call it through fal, where it is strong and weak, the business behind it, and what it means for marketing teams. For more on per-second versus credit pricing, see our breakdown of MixVio’s credit pricing maths.
Table of contents
- What Creatify Boreal Is
- Creatify Boreal Pricing Against Other Video Models
- How Fast Creatify Boreal Generates Video
- How Creatify Boreal Was Built on LTX-2.5
- What the Creatify Boreal Benchmarks Show
- Where Creatify Boreal Is Strongest and Weakest
- Using Creatify Boreal Through the fal API
- The Business Behind Creatify Boreal
- What Creatify Boreal Means for Marketing Teams
- Open Questions About Creatify Boreal
- Frequently Asked Questions About Creatify Boreal
- References
What Creatify Boreal Is
Creatify Boreal is a general-purpose video generation model aimed at advertising work: product shots, user-generated-content style clips, presenter videos and social b-roll. Creatify’s model page describes it as “general video generation built for cost and speed”, with “comparable output for a fraction of the price, with results in seconds”.
A text-to-video and image-to-video model
At launch Creatify Boreal supports two modes. Text-to-video starts from a written prompt. Image-to-video starts from a reference image, such as a product photo or a picture of a creator, and animates it. In both modes the audio is generated together with the pictures, so speech, music and sound effects arrive in the same file.
Where Creatify Boreal is available
The release lists four ways in: the Creatify Labs API, the Model Playground and AdFlow features inside Creatify’s own app, and the fal platform. fal lists the model as creatify/boreal, marked for commercial use. Creatify’s pricing figures are quoted for the fal API.
The second model from Creatify Labs
Creatify Labs calls itself “the research lab building lifelike AI video models”. Its first in-house model, Aurora, is a diffusion transformer that turns a single photo and one audio clip into a lifelike talking video. Aurora reached fal by 9 February 2026. The lab says its team comes from Meta, Snap and FAIR. Its production counter shows more than 100,000 videos generated a month and roughly 465 hours of video in total, without saying over what period the hours were counted.
| Factor | Aurora | Boreal |
|---|---|---|
| Order | First in-house model | Second in-house model |
| Job | Talking avatar from one photo and one audio clip | General text-to-video and image-to-video |
| Audio | Driven by your audio or text to speech | Generated jointly with the video, or driven by your audio |
| Price | $0.07 per second at 480p, $0.14 at 720p | About $0.01 per second |
| Base | Creatify’s own diffusion transformer | Post-trained Lightricks LTX-2.5 |
| Best at | Single-person speaking and singing | Short clips, products, simple scenes |
Creatify Boreal Pricing Against Other Video Models
Price is the first of the two headline claims, and it is the easiest to check because Creatify publishes a comparison table. The difficulty is that three Creatify pages describe the saving three different ways.
One cent per second, billed per second
Creatify Boreal is billed per second of finished video “rather than per credit or token”, the release says, “so the bill for a job can be worked out before it runs”. At one cent a second, a five-second clip costs about $0.05 and a ten-second clip about $0.10. That transparency is useful on its own. Many video tools sell credits whose value changes from model to model.
The fal list-price comparison
The Creatify Boreal model page compares approximate fal list prices for a clip of about five seconds at 720p. We divided each rival’s price by Boreal’s $0.05 to get the multiple.
| Model | Cost per ~5s 720p clip | Type | Multiple of Boreal |
|---|---|---|---|
| Boreal | ~$0.05 | Text or image to video | 1.0x |
| Seedance (720p) | ~$0.60 | Text or image to video | 12.0x |
| Kling 3.0 Pro | ~$0.56 | Text or image to video | 11.2x |
| Veo 3.1 Fast | ~$0.50 | Text to video | 10.0x |
The chart below shows those same list prices in dollars per five-second clip, with the longest bar set to the highest price ($0.60 = 100%).
Three different savings claims
The table supports savings of 10x to 12x. The wording around it varies. The release says a five-second clip costs “6 to 11 times less per clip than the models in Creatify’s benchmark set”. The Creatify Boreal model page says “10x to 47x less per clip”. The Creatify Labs home page says “roughly 10x cheaper than the lowest-cost competitors”. None of the three rival prices in the table produces a 6x or a 47x multiple.
The 47x figure probably reflects a pricier configuration of a rival model that is not in the table, and Creatify’s LinkedIn launch post compares Boreal with Seedance 2.5 on that basis. The Boreal model page also carries a caveat under its prices: “Figures pending final confirmation before launch.” Treat 10x to 12x against the listed 720p prices as the defensible range until Creatify publishes a single set.
| Where Creatify says it | Claim | Matches the price table? |
|---|---|---|
| Launch release | 6 to 11 times less per clip | Partly; table shows 10x to 12x |
| Launch release | 12 times cheaper than a leading closed model | Yes, equals the Seedance row |
| Boreal model page | 10x to 47x less per clip | Lower end only |
| Labs home page | Roughly 10x cheaper than the lowest-cost rivals | Yes, equals the Veo 3.1 Fast row |
| Boreal model page benchmark | Seedance 2.0 costs roughly 30x per second | Different price basis from the table |
How Fast Creatify Boreal Generates Video
Speed is the second headline claim, and Creatify treats it as equal in weight to price. For ad teams that regenerate a shot many times, the wait between attempts can matter more than the bill.
Realtime at one to one
Creatify Boreal “serves at 1:1 with realtime”, the release says, “generating a second of finished video for every second of GPU time”. In practice that means a five-second clip in roughly five seconds. The comparison is with “the minutes-long queues typical of the closed models it is measured against”.
About 40 times the speed
The release puts Boreal at “about 40 times the speed” of a leading closed model. The model page makes the same point against Seedance 2.0, which it says costs “roughly 30× the cost per second and 40× the wait”. A 40x wait on a five-second job would be about 200 seconds, which fits the minutes-long queues the release describes.
The storyboard sum that does not add up
The release gives a worked example. “On Boreal, a storyboard of twenty five-second shots costs about quarter of a dollar, and finishes in under two minutes. The same job on a leading closed model runs roughly $12.” Twenty five-second shots are 100 seconds of video. At one cent a second, that is $1.00, not $0.25. A quarter of a dollar buys 25 seconds of video, or five such shots.
The $1.00 figure is the one consistent with the rest of the release. $12 divided by $1.00 is exactly the “12 times cheaper at the benchmarked price” the release claims, and $12 across 20 shots is $0.60 each, the Seedance row in the table. A quarter of a dollar would imply a 48x saving. It looks like an arithmetic slip, but it is the number most likely to be quoted, so budget on $1.00.
How Creatify Boreal Was Built on LTX-2.5
Creatify Boreal is not a new foundation model. It is an open base model that Creatify has tuned for advertising. That choice explains both its low price and the way Creatify measures its gains.
An open 22-billion-parameter base
Boreal is built on LTX-2.5, “the open-source video generation model from Lightricks”, which the release describes as “a 22-billion-parameter system that generates audio and video jointly”. Lightricks publishes LTX-2.5 22B variants and add-ons on Hugging Face. Creatify says its recipe “will follow the open frontier to newer base models as they mature”.
One post-training recipe on ad footage
Creatify Labs “tuned that base with a single post-training recipe on advertising footage”. This is a form of transfer learning: the base model already knows how to make video, and the extra training shifts it towards what ads need, such as intact product shots and consistent presenters. Creatify has not published the size of the ad dataset or the recipe itself.
Identical arms in every comparison
To isolate the effect of post-training, Creatify says every comparison “holds the two arms identical except the trained weights, using the same prompt, image, resolution and seed”. That is good practice. It means the preference result below measures the recipe, not luck in prompt choice or random seeds.
The same recipe on the previous base
Creatify adds that “running the same recipe on the previous open base moved it the same way”, which it presents as evidence the gain comes from the method, not from a single lucky checkpoint. Lightricks’ earlier LTX-2.3 22B models are on Hugging Face, but Creatify does not name the previous base or give that result’s numbers.
What the Creatify Boreal Benchmarks Show
Creatify publishes two kinds of evidence: a blind human preference test against the untrained base, and an ad-quality pass-rate chart against rival models. Both are useful. Both rest on small samples.
81% of decisive comparisons
“In blind side-by-side review, evaluators preferred Boreal to its untouched base model in 81% of decisive comparisons (25 to 6, with 9 ties, across 40 production cases; sign test p<0.001).” The arithmetic checks out: 25 wins out of 31 decisive cases is 80.6%. Counted across all 40 cases, Creatify Boreal won 25, or 62.5%, and tied or lost 15, or 37.5%.
The ad-quality pass-rate chart
The model page adds a “launch benchmark” of 20 ad prompts. A pass means the model “beats or ties the human-made reference”. Cost is per second of finished video, and Creatify notes that “Kling’s score is calibrated from its direct match-ups vs Boreal and Seedance”.
| Model | Ad-quality pass rate | Cost per second | Passes out of 20 |
|---|---|---|---|
| Seedance 2.0 | 80% | $0.30 | 16 |
| Creatify Boreal | 70% | $0.01 | 14 |
| H3 Max | 65% | $0.08 | 13 |
| Kling 3.0 Pro | 62% | $0.17 | 12.4 (calibrated) |
| Veo 3.1 | 58% | $0.40 | 11.6 (not a whole number) |
| Base LTX-2.5 | 50% | $0.04 | 10 |
Quality per dollar
Creatify’s chart is titled “best ad quality per dollar”. To make that concrete, we divided each pass rate by the cost in cents per second. Creatify Boreal scores 70 pass-rate points per cent. The next best is its own untrained base at 12.5, because the base is listed at four times the price. The chart below scales every bar against Boreal’s 70.
What small samples can and cannot show
With 20 prompts, each prompt moves a pass rate by five points. Creatify Boreal’s 70% lead over H3 Max’s 65% is one prompt, and Seedance 2.0’s lead over Boreal is two. Veo 3.1’s 58% is not a whole number of passes out of 20, and only Kling is flagged as calibrated, so the method behind that bar is unclear. The preference test’s p-value is strong, but 40 cases is still a small, vendor-chosen set.
None of this makes the results wrong. It means the price-per-quality gap is the robust finding, because it is driven by prices that differ 8x to 40x, while the quality ordering among the cheaper models could easily change on a different set of 20 prompts.
Where Creatify Boreal Is Strongest and Weakest
Creatify is unusually direct about the model’s limits. “It’s a new model and we’ll say where it’s weakest,” the model page says.
Products, labels and creators stay consistent
The release argues that for marketers “the deciding factor is often neither price nor speed but consistency”. In Creatify’s report, Creatify Boreal holds “a personalized product label and on-screen name legible for a full clip”, keeps a creator “as one continuous person rather than swapping them mid-shot”, and “renders an app interface legibly where the base model paints a generic wallpaper”. Creatify also reports gains in image quality and fine-detail retention, with less warping.
Short clips and simple scenes
“Short clips and simple scenes are where it’s strongest,” Creatify says. That suits the average clip in its study, which it puts at about five seconds. Longer shots and busy scenes with many moving elements are where teams should test before committing.
Single-person talking clips
The weakest case is a single person talking to camera. There, Creatify says, “the gap to the base narrows most, at 60%”. That is well below the 81% overall preference. It is also the job Aurora was built for, which suggests Creatify expects customers to use the two models side by side rather than one for everything.
| Use case | Creatify’s own assessment | Practical advice |
|---|---|---|
| Product hero shots | Label and product stay intact | Good first test; start from a product photo |
| UGC-style creator clips | Creator stays one continuous person | Check likeness consent for real people |
| App demos | Interface rendered legibly | Verify on-screen text frame by frame |
| Short simple scenes | Strongest area | Ideal for volume variant testing |
| Single-person talking clips | Weakest; 60% preference over base | Try Aurora instead |
Using Creatify Boreal Through the fal API
The fal listing gives the clearest picture of how developers actually call Creatify Boreal. Creatify also documents Boreal endpoints in its own API reference, where jobs are asynchronous: a request returns a job straight away, and you poll for the result or supply a webhook URL.
Structured prompts
fal’s schema asks for “structured directions using [VISUAL], [SPEECH], [SOUNDS], and [TEXT] sections”. Splitting a prompt that way lets a team say separately what the camera sees, what a presenter says, what the soundtrack contains and what on-screen text should read. English is the validated prompt language.
Inputs, duration, resolution and format
| Parameter | Options on fal | Default |
|---|---|---|
| prompt | Required; [VISUAL], [SPEECH], [SOUNDS], [TEXT] | None |
| image_url | Optional reference for identity and composition | None |
| audio_url | Optional driving narration or soundtrack | None |
| negative_prompt | Content to avoid | Empty |
| duration | Seconds, rounded to a legal frame count (5s = 121, 10s = 241, 15s = 361 frames at 24 fps) | 10 seconds |
| resolution | 720p, 1080p or 2k | 720p |
| aspect_ratio | auto, 16:9, 9:16, 1:1, 4:3, 3:4 | auto |
The price comparisons Creatify publishes are for about five seconds at 720p. The pages we read do not give a separate Creatify Boreal price for 1080p or 2k output, so check fal’s billing page before budgeting at higher resolutions.
Provenance labels
fal says the MP4 output includes invisible C2PA metadata and watermarking, with an optional visible “AI-generated” disclosure label. C2PA is an industry standard for recording where media came from and how it was made. For advertisers facing platform rules on synthetic media, having both marks built in is a practical advantage.
The Business Behind Creatify Boreal
Creatify started as an application company, not a model lab. Its move into in-house models says a lot about where margins sit in AI video.
From ad tool to model lab
In mid-2025 Creatify announced a $15.5 million Series A co-led by WndrCo and Kindred Ventures, taking total funding to $23 million. It said it was used by more than a million marketers and 10,000 teams, and had passed $9 million in annual recurring revenue within 18 months of launch. That round coincided with AdMax, which Creatify called an end-to-end AI ad agent.
Why a cheap model matters inside AdFlow
Every clip an ad app generates on a third-party model is a cost the app must pass on or absorb. Running Creatify Boreal inside AdFlow and the Model Playground gives Creatify a generator that its own benchmark places near the top on ad quality at a small fraction of rival list prices. That is our reading of the strategy, not a statement from Creatify, but the release’s focus on volume pricing points the same way.
The open-weights layer
Creatify Boreal also shows a growing pattern in generative video: specialist companies post-training open base models instead of training their own. Lightricks supplies the base, and Creatify supplies advertising data and evaluation. The risk for Creatify is that the edge lasts only as long as its recipe keeps improving newer open bases faster than rivals can copy it. Aggregators already resell many rival models side by side, as our look at Anyvids and the models inside it showed, so a cheap model has to stay good to keep its place in those menus.
What Creatify Boreal Means for Marketing Teams
For most teams the interesting question is not whether Creatify Boreal is the single best video model. It is whether a model this cheap changes how many creative variants are worth producing.
Budget arithmetic for variant testing
Take 20 ad concepts with five variants each, at fal’s default ten seconds: 1,000 seconds of video. At one cent a second, Creatify Boreal would cost about $10. Using the per-clip list prices in Creatify’s table divided by five seconds, the same job would cost about $120 on Seedance at 720p, $112 on Kling 3.0 Pro and $100 on Veo 3.1 Fast. These are list-price estimates, before retries.
Review time becomes the bottleneck
“Every video model gets things wrong sometimes, so people regenerate constantly,” Creatify’s model page says. “Boreal makes that cheap.” When generation costs pennies, the expensive part becomes the human who watches every variant, checks the product and the claims, and approves what runs. Plan review capacity before scaling volume.
Disclosure, rights and brand safety
Cheap generation does not remove legal duties. Image-to-video of a real creator needs that person’s consent, product claims still need substantiation, and ad platforms have their own rules on synthetic media. Keep the C2PA metadata intact through editing, and use the visible label where a platform or regulator expects disclosure. If AI video is new to your team, our AI models and tools hub is a good place to compare options.
Run your own blind test
Creatify’s method is easy to copy. Pick 20 of your own real briefs, generate each with Creatify Boreal and your current model using the same inputs, and have reviewers choose without knowing which is which. Count wins, ties and losses, and add the cost. Your own 20 prompts will tell you more than any vendor’s.
Open Questions About Creatify Boreal
Several things a buyer would want to know are not yet public.
Final prices at every resolution
The model page’s figures are marked as pending final confirmation, and the published comparison covers only about five seconds at 720p. Prices for 1080p and 2k output, and for the Creatify Labs API outside fal, should be confirmed directly.
The full benchmark report
The release refers to “the company’s report”, but the pages we read summarise it rather than publish it in full. The ad dataset, the reviewers, the 20 benchmark prompts and the calibration method for the rival scores would all help buyers judge the numbers.
Licence terms from the base model
Creatify Boreal is a post-trained version of an open model. Buyers using it for commercial ads should ask how Lightricks’ licence for LTX-2.5 flows through to Boreal’s outputs and API terms, particularly for large enterprises.
Longer and more complex clips
Creatify says the model is strongest on short clips and simple scenes. How it behaves at 15 seconds, the longest example duration in fal’s documentation, and in multi-subject scenes is the next thing independent testers will want to see.
Frequently Asked Questions About Creatify Boreal
What is Creatify Boreal?
Creatify Boreal is an AI video model from Creatify Labs, launched on 15 September 2026. It generates video with sound from text or an image, and it is built by post-training Lightricks’ open LTX-2.5 model on advertising footage.
How much does Creatify Boreal cost?
About one cent per second of video on fal, or roughly $0.05 for a five-second clip. Creatify bills per second of finished video rather than in credits. Its figures are for 720p and marked pending final confirmation.
How fast is Creatify Boreal?
Creatify says it runs at one second of video per second of GPU time, so a five-second clip takes about five seconds. It claims about 40 times the speed of a leading closed model.
Is Creatify Boreal better than Seedance or Veo?
On Creatify’s 20-prompt ad benchmark, Seedance 2.0 passed more prompts (80% against 70%), but at roughly 30 times the cost per second. Creatify Boreal passed more than Veo 3.1, Kling 3.0 Pro and H3 Max. The sample is small, so run your own test.
What is the difference between Creatify Boreal and Aurora?
Aurora turns one photo and one audio clip into a talking avatar video, and costs $0.07 to $0.14 per second. Creatify Boreal is a general text-to-video and image-to-video model at about $0.01 per second, weakest on single-person talking clips.
References
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.