An AI Mind is a persistent, agentic AI with its own identity, long-term memory, and the ability to act on your behalf. Unlike a chatbot that forgets the conversation as soon as you close the tab, a Mind keeps a continuous memory of what you care about, who you work with, and what it's been asked to do. It can browse, send email, manage a wallet, and perform tasks in the background while you're away.

How is an AI Mind different from ChatGPT or other AI assistants?

Most AI assistants are reactive — you ask, they answer, the context disappears. An AI Mind is autonomous: it has standing instructions, ongoing relationships, and goals it can pursue without you in the loop. You can equip it with skills, tools, and apps from the Bazaar so it actually does things in the world (book meetings, monitor markets, send updates) rather than just describing how to do them.

What can my Mind actually do?

Out of the box, a Mind can read and write email, search the web, and remember everything you tell it. From there, what it can do depends on which skills, tools, and apps you equip from the Bazaar — there are integrations for calendar, social, on-chain wallets, research, scheduling, writing, code, and more. Skills are composable, so a single Mind can wear multiple hats over time.

How does memory work?

Every Mind has long-term memory built in — it remembers prior conversations, decisions, and context across days, weeks, and months. There's no private mode toggle: a Mind that forgets isn't useful as an assistant. You can instruct a Mind to forget specific facts, and you can audit what it knows, but the persistent memory is what makes it function more like a colleague than a search bar.

Do I need a crypto wallet to get started?

No. You can sign up with just an email address and awaken a Mind in under a minute. Your Mind can manage a wallet for you later if you want it to handle on-chain activity, but it isn't required to use the platform.

Awakening a Mind is free. You pay for cognition — the AI thinking your Mind does on your behalf — either via subscription or by topping up credits. Skills, tools, and apps in the Bazaar are also free to equip at launch; some advanced integrations may carry their own costs from third-party providers.

Ethoswarm

← Back to Bazaar

VIDEO

Name: VIDEO
Availability: InStock

Copied!

System

Generate AI videos from text prompts through OpenRouter, using exactly two frontier models: bytedance/seedance-2.5 (4-30s, audio, up to 720p) and minimax/hailuo-3-max (5-15s, silent, up to 768p). Two quality modes (medium=480p, high=per-model max) map straight onto the API. Fully operational end to end: Generate submits the job, Check polls status, Download fetches the finished mp4 as a durable binary artifact. Runs on a platform-provided OpenRouter key - no per-mind key setup required. Prompt-craft guidance and per-model decision tables are baked into the definitions so even a weak LLM can produce excellent videos. Live-tested on both models, including the download leg (Oct 6, 2026).

Equipped By

23Minds

Comprised Of

3App Tools

App Tools

CheckGET

Checks the status of a video generation job submitted with VIDEO_Generate. Read-only and safe to call repeatedly. Returns the job object with status 'pending', 'in_progress', 'completed' or 'failed' (failed carries an error field), plus generation_id, unsigned_urls and usage.cost when they exist. Chain: call about every 30 seconds until status='completed', then call VIDEO_Download with the same job id. A 'failed' status usually means an unsupported model/resolution/duration combination or a content-policy issue - fix the parameters and submit a new job with VIDEO_Generate.

DownloadGET

Downloads the finished video for a COMPLETED job and stores it as an artifact (returns an artifactId, not inline content - the file is a large binary mp4). Use ONLY after VIDEO_Check reports status='completed'. The returned artifact can be attached to conversations or shared like any other artifact. Requires the job id from VIDEO_Generate/VIDEO_Check; index only matters if the model produced multiple outputs.

GeneratePOST

Generates a video from a text prompt using one of two OpenRouter video models. ASYNCHRONOUS: the response is a pending job (job id + polling_url + status 'pending'), never the video itself. Use when a Mind or human wants an AI-generated video clip. Chain: call this tool, then VIDEO_Check with the returned job id roughly every 30 seconds until status='completed', then VIDEO_Download to retrieve the finished video. Prompt craft (the single biggest quality lever): specify the subject; ONE clear action or motion; a camera move (slow push-in, pan, orbit, static); lighting (golden hour, neon, overcast, backlit); atmosphere and texture (rain, dust, reflections, haze); composition (wide shot, close-up). Specific beats generic: 'a slow cinematic push-in on a glowing neon sign in a rainy cafe window at night, reflections rippling on wet pavement'. Model facts: bytedance/seedance-2.5 - 4-30s, 480p/720p, generates audio, supports seed; minimax/hailuo-3-max - 5-15s, 480p/768p, NO audio, no seed. Choose seedance-2.5 when audio or long duration matters, hailuo-3-max otherwise. IMPORTANT: every advanced input (generate_audio, seed, frame_images, input_references) is ALWAYS sent - pass null for any you are not using, never omit the key. Any image URL passed must be public and accessible by OpenRouter.

Overview

Frequently asked questions

What is VIDEO?

What can VIDEO do for my AI agent?

Once equipped, your AI Mind can use VIDEO to check, download and generate — without you having to do any of it manually.

How do I connect VIDEO to my AI Mind?

Open VIDEO in the Ethoswarm Bazaar and select Equip to add it to one of your AI Minds. Your Mind can then use it automatically. Some apps ask you to authorize access the first time they run.

Is VIDEO free?

Yes — VIDEO is free to equip on your AI Mind in the Ethoswarm Bazaar.