Muse Image
Meta · AI imageMuse Image is an agentic image generation model from Meta, served on VidMachine through OpenRouter's dedicated Images API. Unlike single-pass models, it reasons about the prompt before rendering, then produces the image from text and optional reference images.
On VidMachine it is billed at 2 credits per generated image, making it one of the lowest-cost models for scene stills and start frames.
Key features and benefits
Reasoning before rendering
Muse Image is designed as an agentic model: it works through the request before producing pixels, which helps with complex scenes, composition instructions, and prompt details that single-pass models often miss.
Text-to-image and editing
Generate an image from a prompt alone, or attach reference images to guide style, character consistency, and edits. VidMachine passes project and previous-segment references when available.
Aspect ratios
Muse Image renders a limited set of native shapes: 1:1, 2:3 (portrait), and 3:2 (landscape). VidMachine requests the closest supported shape and center-crops to the exact project format, so vertical projects get a true 9:16 (1080×1920) frame and landscape projects get 16:9 (1920×1080).
OpenRouter routing
On VidMachine this model runs only through OpenRouter's dedicated Images API (POST /api/v1/images), not Replicate. You need OPENROUTER_API_KEY configured in the environment; billing uses VidMachine credits per image.
Low credit cost
At 2 credits per image, Muse Image is a cost-efficient option for high-volume scene generation and iteration alongside Pruna AI's P-Image.
Technical specifications
ProviderMeta
OpenRouter modelmeta/muse-image
InputsText and reference images
OutputImage
Native aspect ratios1:1, 2:3, 3:2
Credits2 per image
VidMachine routingOpenRouter Images API (POST /api/v1/images)
Use cases and applications
Muse Image suits automated shorts pipelines—starting frames for AI Video, full scene stills for AI Images, and high-volume iteration where per-image cost matters.
Use it when a scene prompt has layered composition or detail requirements and you want the model to reason about the request before rendering.
Why this model
Choose Muse Image for Meta-backed, agentic image generation at a flat 2 credits per image through OpenRouter's Images API.
How VidMachine uses it
Internal model id: muse-image. Generation uses OpenRouter POST /api/v1/images. Because Muse only renders 1:1 / 2:3 / 3:2, VidMachine requests the closest supported shape (2:3 for portrait, 3:2 for landscape) and then center-crops to the exact project format—9:16 (1080×1920) or 16:9 (1920×1080).
Reference images from the project and previous segment frames are attached as input_references when available.
Billed at 2 credits per generated image.
What you should know
Does Muse Image run on Replicate in VidMachine?
No. This model is OpenRouter-only on VidMachine. Select muse-image in image_model_priority and ensure OPENROUTER_API_KEY is set.
Can Muse Image generate exact 9:16 or 16:9 images?
Not natively. Muse renders 1:1, 2:3, and 3:2 only. VidMachine asks for the closest shape and automatically center-crops the result to a true 9:16 or 16:9, so the final asset always matches your project orientation.
How much does Muse Image cost?
2 credits per generated image on VidMachine.
Does it support reference images?
Yes. Muse Image accepts reference images alongside a text prompt for guided generation and editing; VidMachine passes available project and segment references.