AI Render Critic

The Render Critic sends a render or viewport capture to your AI provider's vision model and returns a structured critique — a fast way to get a second opinion on lighting, composition, and materials without leaving Blender.

Requirements

Requires a vision-capable provider: Claude, Gemini, or GPT-4o-mini. Groq and Ollama are not used for critiques even if selected as your chat provider, since the critic needs image understanding.

How It Works

1. Trigger a critique from the Chat tab (either by asking directly, e.g. "critique this render", or via the dedicated critic control).
2. BlenderMCP Pro renders the current frame to a temporary file, reads it back, base64-encodes it, and discards the temp file.
3. The image is sent to your configured provider along with a structured critique instruction.
4. The response comes back with an overall quality score out of 10 plus targeted feedback, and is displayed in the chat panel.

Focus Modes

You can steer the critique toward a specific concern instead of a general pass:

• Full — a comprehensive review covering everything below.
• Lighting — quality, shadows, highlights, and overall mood.
• Composition — framing, balance, focal points, and visual flow.
• Materials — surface realism, shading, and texture quality.
• Technical — noise, artifacts, clipping, and general render quality issues.

Viewport Capture vs. Full Render

The Render Critic uses your scene's actual render output (via render_to_file-equivalent internal logic), so it reflects your real render engine, samples, and settings — not just a viewport preview. If you want a quick check without waiting for a full render, use the separate viewport screenshot feature in the chat panel instead (see AI Chat Panel & Providers), which captures the 3D viewport directly and is much faster, at the cost of not reflecting your final render settings.