Images: GPT Image 2.5, Nano Banana 2, Seedream 5.0, Qwen Image 3, Grok Imagine
Thirteen image models, from quick edits to dense infographics. On a paid account a new image starts on GPT Image 2.5 Flare unless you name another.
SayMaker AI Video Agent
Type what you want to see, in any language. The AI video agent turns one rough sentence into a detailed prompt, picks the image, edit, clip or animation step on any of 26 models, and keeps shaping the result in one thread until the clip on screen is the one you could only picture before.
Picks the stepStill, then motionFree account, free credits
Ideas to try
How it works
How the AI video agent works
Four moves: you set the direction, the agent plans the steps and their credits, and short follow-ups steer the result.
Say what you want
Write it the way you would brief a person: "a shiba on a neon motorbike, rainy night, then make it drive toward the camera". Attach photos if the real subject matters.
It chooses the tool
The agent reads intent and calls one of four tools: generate an image, edit an image, generate a video, or animate an image you already have. It can use any model the generator offers.
It shows the plan first
Any video, and any job with more than one step, starts as a plan card: each step, its model, its credits, the total and your balance. Nothing is spent until you confirm, and the run must match what you approved.
Keep talking
"Warmer light." "Put her in the jacket from this photo." "Now ten seconds." Every earlier result stays in the thread, so short follow-ups work. Video renders in the background while you keep talking, and the clip appears in its card when it is done.
One request, several steps
Jobs that used to take three tools
A generator does one step per click. The AI video agent strings the steps together, and each one starts from the previous result instead of from scratch.
Image → video
Make the frame, then move it
Ask for a product shot, approve it, and write "animate it, slow push-in". The agent passes that exact picture to a video model, so the clip opens on the frame you already liked.
Make my frame movePhotos → composite
Combine people, clothes and logos
Attach a portrait and a jacket and say "have him wear this". The agent sends both photos into one edit and labels them image 1 and image 2, so neither gets dropped.
See Nano Banana 2Marked-up edit
Draw on the photo instead of describing it
Circle the lamp, draw an arrow at the window, and send. The marks tell the agent where to work; they never show up in the final image.
See GPT Image 2.5Models
The models the AI video agent runs
The AI video agent runs the same 26 models as SayMaker's generators, at the same prices. Name one in your message or pick it in the composer; leave it out and the agent uses the default for your account.
Video: Veo 3.1, Kling 3.0, Seedance 2.5, Wan 3.0 and more
Thirteen video models, all from the same AI video agent conversation: Veo 3.1 as the paid default, Wan 3.0 or Seedance 2.5 for takes up to 30 seconds, and Kling 3.0 for heavy motion in 4K.
Free accounts: SayMaker models and Nano Banana 2 Lite
Without a plan the agent uses SayMaker Image v1 for new pictures, Nano Banana 2 Lite for photo edits and SayMaker Video v1 for clips. Paid models are listed with a free alternative.
Who uses it
What people ask an AI video agent to do
Bring the idea you cannot stop picturing, a hook, an ad, a dance, a story beat, and leave with a finished clip built in one AI video agent conversation instead of a pile of half-learned tools.
Explainer clips
A key frame for the hook, then the same scene animated with sound for the opening seconds of the video.
CreatorsJoke and meme posts
An absurd still first, then a short loop of it for the platforms that reward motion.
SocialPerformance and dance
Seedance 2.5 or Veo 3.1 when the clip needs sound baked in, picked by name in the same message.
MusicProduct ads
The real product photo edited into a new setting, then a slow camera move for an ad slot.
ShopsCharacter sheets
One character, redrawn in new poses by follow-up messages, with earlier versions kept in the thread.
IllustratorsStory beats
Frame by frame through a scene, animating only the shots that earn it, so the budget goes where the motion is.
FilmmakersAnswers
AI video agent FAQ
Straight answers about what the agent does, what it costs and how to get the most from it.
An AI video agent takes a plain-language request and decides which steps and models it needs, instead of asking you to pick a tool and fill in its settings. SayMaker's agent can generate an image, edit one, generate a video, or animate an image, and it can run those in sequence inside one conversation.
Yes. Ask for a clip and the SayMaker agent plans it, shows the model and credits, and after you confirm it renders in the background while the chat carries on. A paid account defaults to Veo 3.1; free accounts use SayMaker Video v1; any other video model runs when you name it on a plan or credit pack.
Yes. A free account is all it takes: signing up adds free credits, so the agent starts making images and clips on the free models right away. Paid models such as Veo 3.1, Kling 3.0 and Seedance 2.5 open with a plan or credit pack, and runs that fail for a technical reason are refunded.
Yes, many of the video models write sound into the clip, including Veo 3.1, Kling 3.0, Seedance 2.5, Wan 3.0 and LTX 2.5. Kling 3.0 Turbo stays silent for footage under your own audio. Ask for sound and name a model that has it, or let the plan card show which model the agent picked.
Yes, once you are signed in. Attach one or several photos and describe the change. You can also draw circles and arrows on an image to point at what should change, and the marks stay out of the result.
Any language you are comfortable with. The agent replies in the language you used, and it turns even a one-line image request into a detailed English prompt behind the scenes, adding lighting, style and composition without changing what you asked for.
The agent explains what happened and what to change next, such as a different source photo or a new direction for the edit. It retries a failing run once at most and stops the moment inputs are refused, so your credits are protected, and runs that fail for a technical reason are refunded.
Video costs many times more than an image, and a multi-step job adds up. The plan card lists each step with its model and credits, the total and your balance, so you approve the spend before it happens. If the agent tried to run a dearer model than the one you confirmed, the run would be refused.
A generator page runs one model with the settings you choose. The agent takes over jobs with several steps, carrying each result into the next so you describe the result instead of operating each tool. Every model page on SayMaker, like Veo 3.1, still has its own generator.
Yes: make it once and share it anywhere by switching the conversation from only-me to public, which gives it a link anyone can open. Images from all your chats also collect in a library, so earlier results are easy to find.
Describe the clip. Let the agent handle the steps.
Sign up free and one message is enough to start the AI video agent. Nothing is spent until you approve the plan, and runs that fail for a technical reason are refunded.