ConceptsMedia
Images, video and sound
Media is a job like any other. Ask for an image, a clip or a sound; Swarm sends it to a model you connected, keeps the files with the job and saves them into your project.
Three kinds
Product shots, illustrations, banners, icons. Default model: Nano Banana 2.1.
Teasers, demos, social clips. Default models: Gemini Omni 1.1 Flash or Veo 3.1.
Jingles, voice-overs and effects, from the sound models you connect.
Which models are ready depends on what you connected: your ChatGPT plan, a Google key or OpenRouter. Your AI can list them with their options and prices.
Create, edit, extend
From a description: “a calm product shot of the mug on oak, morning light.”
Change an earlier result or an image from your project: “make the background white.”
Carry a clip on: “another 8 seconds, the camera pulls back.”
Options
Every request is checked against the model before anything is spent, so an impossible size or length fails at once instead of costing money.
| Option | What it sets |
|---|---|
| Aspect ratio | 16:9, 9:16, 1:1 and whatever else the model offers. |
| Resolution | Up to what the model supports, for example 1080p video. |
| Length | Seconds of video or sound. |
| Count | Several variations in one job. |
| Inputs | An earlier result, or an image from your project, to edit or extend. |
| Model | Any ready model of that kind; otherwise the default for the kind. |
Where the files go
Each media job keeps its files. When you're happy, your AI saves them into a folder of your project; it never overwrites a file that's already there. Media jobs also show up in your runs and your spend like every other job.
Things to say
Make three square product shots of the new mug for the shop.
swarm_media_modelsswarm_request_media
Here are three. The second has the warmest light. Save it to web/public/img?
Turn the second one into a 10-second vertical video for Instagram.
swarm_request_media
Every Monday, make a banner for this week's sale.
swarm_deploy_workerswarm_schedule_job
For your AI
swarm_media_models lists each ready model with its ratios, resolutions, lengths, counts, inputs, operations and pricing. swarm_request_media makes, edits or extends; swarm_save_media writes the files into a project folder. Details: ai/tools.md.