Ares-II is not a model. It is the orchestration layer that sits above multiple AI models and manages the entire production pipeline. It plans, generates, evaluates, and refines on your behalf.
Every request flows through four specialized agents. Each handles a distinct phase of the production process.
Generates, inspects its own output visually, and re-runs if quality falls short.
Internal knowledge files per model covering capabilities, parameters, and strengths. Only relevant context loads per task.
Switches between models mid-pipeline based on what each step requires. No manual selection needed.
Every task comes with a shape — aspect ratio, duration, reference frames, continuity constraints. Ares-II scores candidate models against that shape and picks the best fit. You can override, but by default the engine routes itself.
Ares-II has access to 50+ AI models across 10+ providers — covering image generation, video production, voice, audio, and utility tasks. It selects the best model for each task automatically. Model availability expands continuously.
Ares-II runs the same pipeline from either surface. Drive it visually on the studio canvas, or drive it conversationally from the chat — the agent picks the model, calls the tool, and renders the result in both.
Node-based canvas for multi-scene storyboards, character management, and shot-level control. Connect prompts to scenes, scenes to image and video slots — the pipeline is visible at every step.
Describe what you want in natural language. Reference canvas nodes by name, override a model when you need to, and let Ares-II plan, run the tool, and stream the result back inline.
Every project in Studio is backed by SDL, a structured description language that represents your scenes, shots, characters, and media as code. Ares-II reads and writes this structure directly — it understands your project the same way the system does. When you ask it to add a scene or adjust a shot, it modifies the underlying project files and the canvas updates automatically.
Scenes, shots, characters, and media are all represented in a consistent format that both you and the AI can work with.
Your project is a set of files. Ares-II can list, read, write, and organize them just like working in a codebase.
Changes made by the AI compile into the visual canvas instantly. What the agent writes is what you see.
The questions people ask before they hand a shot to the agent.
No. Ares-II is the orchestration layer — it plans the work, picks from 50+ candidate models per task, runs the tool, checks the result, and iterates. Models come and go; the agent stays.
Yes. Every shot accepts an explicit model override. Default is auto-routing, but if you want Kling v2 for a specific motion or Veo 3.1 for a wide establishing, you can pin it.
The critic stage runs vision feedback on the generated output — checks composition, continuity, artifact density, subject fidelity — and scores it. If the score falls below threshold, the refiner re-runs with adjusted parameters.
Yes. Ares-II reads the canvas directly — it can see all nodes, understand connections, inspect shots, and write new ones. Anything you do visually it can do in code.
SDL is the structured description language Votocon uses to represent a project — scenes, shots, characters, and references are all expressed as code. Ares-II modifies SDL files directly; the canvas compiles from the same source.
Depends on the task. Simple still images typically resolve in one pass. Complex shots with continuity constraints may go through 2–3 refine loops before the critic accepts.