The AI video generator turns a sentence or a photo into a clip. Pick the model yourself, or let the agent choose.
Long single takes from a prompt or photo, 480p–1080p, billed by the second it runs.
Generate video from text or images. 480p–1080p, with audio.
Cinematic video. Lite and Fast tiers, 720p–4K.
Up to 4K from words or a still; about 320 credits for a silent 6s 720p.
One setting, 2K, so you only choose the length — 5 to 10 seconds.
The cheapest Seedance 2 tier. 480p or 720p video, priced per second.
Seedance 2 between Mini and full; about 745 credits for 6s 720p from a prompt.
Kling 3.0 tuned for speed, 720p or 1080p, billed by the second.
Lightricks' open model: sound and picture in sync, as far as 4K, 20 seconds and 50fps.
Alibaba's Wan 3.0 keeps one shot rolling for 30 seconds, 480p–1080p, with its own soundtrack.
Every major video model worth naming lives in one place: the SayMaker AI video generator. The generator above runs every video model the site carries: Veo 3.1, Kling 3.0 and Kling 3.0 Turbo, Seedance 2.5, Seedance 2.0 with its Fast and Mini tiers, Wan 3.0, MiniMax H3, LTX 2.5 and SayMaker Video v1, plus Seedance 1.5 Pro and Kling 2.6 for even more ways to match the shot.
Each model has a different strength, and the settings change with it. Veo 3.1 prices a clip the same whether it runs 4 or 8 seconds. Wan 3.0 and Seedance 2.5 hold one continuous take for up to 30 seconds. Kling 3.0 is the one for heavy motion. MiniMax H3 renders everything at 2K, and LTX 2.5 runs from a 2-second loop to a 20-second shot.
You do not have to learn all of that before starting. Describe the clip to the agent and it picks a model that fits the job and your account, shows the plan with its credits, and renders after you confirm. Or skip the agent, choose the model yourself in the AI video generator, set length and resolution, and generate.
Every clip draws on the same credit balance as images, and the price for the exact model, length and resolution is on screen before anything renders. A run that fails for a technical reason gives its credits back, and every finished clip stays in your account, ready to become the starting point for the next shot.
You and your person in the orange booth, trading bars at one mic.
One photo, a smile that breaks into tears, then you turn and walk away.
One photo, a bob wig, a 70s suit and the stiff knee-kick shuffle.
You on the dock, your pet singing to camera, both of you at sunset.
Upload a clip, keep every move and camera path, recast the whole cast.
Write the subject, motion, camera and sound in one prompt and get a clip back on any of the AI video generator's 13 models.
Upload a photo, or make the starting frame right in the agent and just say "animate it": two words, and the clip opens on that exact frame.
Veo 3.1, Kling 3.0, the Seedance models, Wan 3.0 and LTX 2.5 generate audio with the picture.
Up to 30 seconds on Wan 3.0 and Seedance 2.5, up to 4K on Veo 3.1, Kling 3.0 and LTX 2.5.
Google's model and the default on a paid account. One price per clip at each resolution, from 115 credits at 720p, with sound.
Kuaishou's model for complex movement, with audio and a native 4K tier. Kling 3.0 Turbo is the silent, faster version for footage going under a voiceover.
Both run a single take of up to 30 seconds with sound. Wan 3.0 costs far less per clip; Seedance 2.5 carries ByteDance's prompt-following.
The Seedance 2 look and sound at 480p and 720p: burst takes on Mini, publish on Fast.
Every clip renders at 2K and the price depends only on length, so the draft you approve is already full size.
Lightricks' open-weights model with audio, from 2 to 20 seconds and 720p to 4K, priced by the second.
The video model free accounts run, from a prompt or a photo, 4 to 15 seconds long. One daily check-in pays for a 4-second clip, so your prompt comes alive before you spend a cent.
Choose from the picker, or describe the clip to the agent and let it choose.
Subject, action, camera, light and sound in one clear scene, or upload a starting frame.
The options change with the model, in the AI video generator or the agent; each combination shows its credits.
Confirm the figure, render, and download the clip or send it on to the next step.
The SayMaker AI video generator runs 13 video models, plus an agent that runs it for you: pick the model here, or describe the result and let it plan the steps and credits.
Thirteen, from Veo 3.1, Kling 3.0 and Seedance 2.5 to Wan 3.0 and LTX 2.5; the full list is in the AI video generator picker above.
On a paid account a video request that names no model runs on Veo 3.1. Free accounts use MiniMax H3 Fast. You can always name another model in your message.
Yes, on Veo 3.1, Kling 3.0, the Seedance models, Wan 3.0 and LTX 2.5, where sound is written in the same render. Kling 3.0 Turbo is silent by design, for footage going under your own audio.
Up to 30 seconds in one take on Wan 3.0 and Seedance 2.5, and up to 20 on LTX 2.5; most others run 4 to 10.
In the AI video generator, most models bill per second by resolution; Veo 3.1 bills per clip, MiniMax H3 by length. You see the credits first, and a technical failure is refunded.
New accounts get free credits for the AI video generator, and every daily check-in adds enough for another MiniMax H3 Fast clip. Premium models such as Veo 3.1, Kling 3.0 and Seedance 2.5 need a plan or credit pack.
Yes. Upload a photo in image to video, or make the image in the agent and say "animate it".
Use the AI video generator when you know the model and settings. Use the agent when a job has several steps, or when you would rather describe the result than compare models.
Describe one shot: who is in it, what happens, where the camera moves, the light and what should be heard.
Video generation is not deterministic, so every run is a fresh take. Use each run to find the prompt that brings the motion you want, then run that prompt at full size.
Most models offer 16:9, 9:16 and 1:1; Veo 3.1 and LTX 2.5 offer 16:9 and 9:16.
A nature-doc shot, an illustration drawing itself, a minimal animation, and a camera move through a harbour city. All four came out of the SayMaker generator, and yours will come out of the same box.
Man Escapes Monsters Through Portal
Man Summoning Green Energy Orb
Man Transforms Into White Werewolf
Man Walking With Red Lines
Tired Man Watching Boy Use Laptop
Mech Winged Ascent
Mechanical Bird Delivers a Cookie
Medieval Soldiers Stand in Formation
Credits are shared across everything SayMaker runs. You see the cost before each run, and a technical failure is refunded.
Checkout runs on Stripe over an encrypted connection. Your card details are handled by Stripe and are never stored by SayMaker.
Take the settings yourself, or hand the job to the agent and approve its plan; a technical failure refunds the run either way.
Create my clip