Fresh generation
Maximum length means one new generation. Extensions, stitched clips and a project's total timeline do not count.
The screening program · October 2026
All seven tools from my video, with official links, example videos, maximum clip lengths, resolution details and API pricing.
The order is my editorial ranking. Prices and availability change quickly, so every volatile claim is dated and linked to an official source.
Five-second verdict
How to read the numbers
Maximum length means one new generation. Extensions, stitched clips and a project's total timeline do not count.
Every dollar figure is an example API rate, not a consumer subscription. The output resolution and important conditions stay attached to the number.
Base or native generation resolution is different from an upscaled or enhanced delivery file. A showcase file's dimensions do not prove the model rendered natively at that size.
These are Matt's picks, informed by official examples and practical workflow fit. No matched independent physics test of these exact seven versions was conducted.

My overall / physics pick
My top pick for advanced physics, complex motion and overall control, especially when you have the budget.
Open Seedance in DreaminaOpens Dreamina through Matt's referral link. The API prices here come from BytePlus ModelArk, separate from Dreamina's consumer plans.
Practical advantage
Up to 30 image, 10 video and 10 audio references, plus green-screen, clay-render and targeted-editing workflows.
Motion / physics note
Matt's advanced-physics favorite. Official examples show cloth, environment-dependent motion and coherent spatial lighting, but this is an editorial pick rather than a controlled benchmark.
Official example · ByteDance Seed
Peking opera motion
A vendor showcase of coordinated costume, prop and body motion.

From 1¢/sec at 480p — October promo
The budget pick for short, simple shots, with dialogue, ambience and effects generated alongside the scene.
Explore HeyGen VideoThis is HeyGen's general-purpose video model, not its avatar engine or the original MiniMax H3 endpoint.
Practical advantage
The lowest verified entry rate among the six tools here with public pricing. Text-, image- and reference-to-video all include generated sound.
Motion / physics note
Best suited to one subject, one place and one contained action. HeyGen's own guidance calls out close handwork and soft motion such as hair, paper and petals as weaker areas.
Official example · HeyGen
Racing-car onboard shot
A vendor example showing a short, contained scene with generated sound.
A strong option for high action, complex camera movement and longer sequences with keyframe control.
Official Kling 4.0 pageThe official model page is the safest access/status reference while API pricing and higher-resolution rollout remain in motion.
Practical advantage
Up to 10 flexible keyframes and 15 combined references for subjects, scenes, motion, voice and camera direction.
Motion / physics note
Official examples show ambitious action, continuity and camera control. Stable motion alone does not prove physically correct contact or conservation.
Official example · Kling AI
Martial arts
An official showcase of fast action and camera movement, not a controlled physics test.

A useful balance of long scenes, style and action, with broad reference inputs.
Explore Wan 3.0The console page may require sign-in; the public documentation link below carries the specifications and rates.
Practical advantage
Images, video, audio, files and links can serve as references for narratives, dialogue, products and recurring subjects.
Motion / physics note
Official sparring and dojo examples show ambitious choreography and continuity. They are showcases, not a matched physics benchmark.
Official example · Alibaba Cloud Model Studio
Sparring in ruins
An official action showcase demonstrating long-scene continuity.
Beautiful output with native 1080p and useful image, voice and keyframe controls.
Open GrokGrok is the product entry point, not a promise that every account automatically has this exact model.
Practical advantage
Up to 14 image and three voice references, plus first, last and up to four intermediate keyframes.
Motion / physics note
Matt rates the physics as decent. xAI reports better weight, momentum and fewer warps than its prior model, not a measured win over the other six.
Official workflow demo · Grok
Grok Imagine Video 1.5 official launch film
A vendor-edited launch reel containing multiple examples; it is not a single fresh generation.
Multi-shot scenes and a draft-to-final workflow that preserves the chosen scene and motion.
Explore FLUX 3 VideoThe official model page links to the BFL playground and API; higher delivery tiers are priced separately.
Practical advantage
Multi-shot composition, native audio, typography, multilingual dialogue and controlled iteration from a cheap draft into the selected final render.
Motion / physics note
The practical differentiator is preserving the chosen scene and motion from draft to final. Broad preference results are not a dedicated physics score.
Official example · Black Forest Labs
Running horses
An official animal-motion showcase from Black Forest Labs.
Cinematic generation with conversational editing and useful first- and last-frame controls.
Official Gemini Omni docsThis is the official model and documentation page, not a direct consumer generator interface.
Practical advantage
Direct and revise a scene conversationally, control first and last frames, and use faster 360p drafts before higher-resolution delivery.
Motion / physics note
Matt likes its cinematic output and physics understanding. Google's world-knowledge positioning is not an independently measured head-to-head win.
Official workflow demo · Google
First- and last-frame interpolation
A workflow demonstration; its edited length is not the maximum for one fresh generation.
The admission board
These are headline examples, not a synthetic price winner. Resolutions, regions, promotions and billing paths differ.
| # | Tool | USD API example | Condition |
|---|---|---|---|
| 01 | Dreamina / Seedance 2.5 | Approx. $0.231/sec | 720p API example |
| 02 | HeyGen Video | From $0.01/sec | 480p — October promotion |
| 03 | Kling 4.0 Flagship | Not publicly verified | Kling 4.0 API |
| 04 | Wan 3.0 Standard | Approx. $0.082513/sec | 720p global rate |
| 05 | Grok Imagine Video 1.5 | $0.14/sec | 720p API output |
| 06 | FLUX 3 Video | $0.17/sec | Standard HD, audio included |
| 07 | Google Gemini Omni 1.1 Flash | Approx. $0.10/sec | 720p video output |
720p is a useful reference point, not an identical quality setting across vendors. HeyGen's row is promotional, Wan is region-dependent, FLUX separates drafts from final renders, Gemini adds input and text/thinking charges, and Kling 4.0 has no verified public rate.
Work with Matt
Book a strategy hour with Matt to match the model, access path, budget and production workflow to the videos you actually need to make.
Want the longer, evergreen reference? Read Matt's Top AI Video Generators guide. This dated page preserves the ranking from the October 2026 video. Read the text version.
The order is Matt's editorial judgment across output quality, motion, control, practical workflow and value. Specifications, access and prices were checked against official vendor pages on October 18, 2026. Official showcases help explain capabilities, but they are not matched laboratory tests and their file dimensions do not prove native model resolution.
#1 pricing and availability
Token-based BytePlus estimate for a 16:9 output without video input. Aspect ratio, dimensions and reference-video input change the total.
#2 pricing and availability
The October 2026 launch promotion is 50% off. The standard 480p rate is $0.02 per second.
#3 pricing and availability
The developer pricing page exposes purchase options after client-side rendering, but a public Kling 4.0 per-second rate was not verified. Unknown does not mean free.
#4 pricing and availability
Rate shown for global regions including US, Frankfurt, Tokyo and Hong Kong. Singapore is $0.10/sec at 720p.
#5 pricing and availability
The API also charges $0.08/sec at 480p and $0.25/sec at 1080p. Extra input images are $0.01 each.
#6 pricing and availability
BFL's current official rate is $0.17/sec for text- or image-to-video at HD. Draft HD is $0.06/sec and the final enhanced render is billed separately.
#7 pricing and availability
Calculated from $17.50 per million video-output tokens at 5,792 tokens per 720p second. Input and text/thinking output are additional.