Why do model options differ on Martini?
Short answer
Each AI model is built by a different provider with different capabilities. One supports 16:9 only, another supports nine aspect ratios. One accepts a reference image, another doesn't. The options shown on each model node reflect what that specific model actually supports — Martini doesn't hide options that work, and doesn't show options that don't.
Input modes vary by model
Some video models support text-to-video only — you give a prompt and they generate from scratch. Others support image-to-video — you give a starting image and the model animates it. Others support video-to-video — you give a source clip and the model restyles or extends it. A few support all three. The node shows only the inputs the active model accepts; switching models in the picker rebuilds the input panel to match the new model's modes.
If a connection on the canvas suddenly turns red after switching models, you likely connected an input the new model does not accept. Check the new model's supported inputs in the node and either remove the unsupported connection or switch back to a sibling model that accepts it. The xyflow layer enforces these constraints to prevent invalid generations.
Aspect ratio, resolution, and duration
Aspect ratio support is provider-specific. Sora 2 supports a wide list of aspect ratios for vertical, square, and cinematic framing. Hailuo offers a smaller set, often capped at 16:9 and 9:16. The aspect-ratio dropdown on the node lists only the options the active model produces — the model picker on Martini does not show options the model cannot generate.
Resolution and duration vary by model and plan. A model may produce 720p on the Standard plan and 1080p only on Pro. Maximum duration may step from 4 seconds to 8 seconds across providers, and longer durations cost proportionally more credits. The node panel shows the current model's supported resolutions and durations, with grayed-out options the active model does not support and lock icons on options gated by plan tier.
Plan-tier gates on options
Some options sit behind plan gates within the same model. A model may produce 720p on every plan but require Pro for 1080p, or accept up to 4-second durations on Free and unlock 8-second durations on Standard+. The lock icon and the Requires {plan} label on the option indicate the gate. See the why-upgrade-prompt-shows article for the full set of plan-gated features and what to do.
Provider-side feature flags can also restrict options temporarily. A model in beta release may unlock a new aspect ratio or duration only for users on a specific tier during the rollout window. Once the feature is generally available, the gate is removed and the option appears for everyone.
Style and parameter differences
Beyond inputs and aspect ratios, models expose different style controls. Some have a motion-strength slider, some have a camera-control panel, some accept a style reference image, some accept a negative prompt. These reflect what the model actually exposes via its API. Martini wraps each provider's API surface — when a provider does not expose a parameter, Martini cannot show it.
If a parameter you used on Model A is missing on Model B, the equivalent control may be expressed differently — a motion-strength slider on one model might map to a separate motion preset on another. Read the model card and try the closest control on the new model rather than expecting a one-to-one map.
Examples
- Sora 2 offers nine aspect ratios; Hailuo offers two — the dropdown reflects each model.
- Image-to-video accepted on Seedance 2 but not on a text-to-video-only model — the input panel changes.
- 1080p available on Pro+ for one model; 720p only on a Standard plan — the lock icon shows.
- Camera-control panel exposed on one model and absent on another — the model decides the controls.
- Maximum duration steps from 4s to 8s across providers — the slider range changes per model.
Edge cases
- A red connection on the canvas after switching models indicates an input the new model rejects — remove or revert.
- Beta features may appear on one tier during rollout and broaden later — check release notes.
- Some models hide options when an upstream input is incompatible (e.g., negative prompt grayed out without a base image).
- Workspace admins can disable specific models or providers — options gated this way appear missing rather than locked.
What to do next
- Open the node's model picker to see which models support the input modes you need.
- Read the model card for each option you don't recognize — it explains the parameter in context.
- Switch to a sibling model if the active model lacks the option you need.
- See the why-upgrade-prompt-shows article if the option is plan-gated on your current tier.
Related help articles
Related models and tools
Still need help? Contact support.