Grida CLI models and voices

Find a supported operation before preparing its input. GG listing and inspection read the current public Gateway catalogue and operation schemas without credentials. Explicit BYOK listing and inspection use the installed adapter inventory and work offline. An unfiltered list combines both, so it needs a connection; select a BYOK provider to browse offline.

A model is shown only when this CLI supports its interface and transport. New GG models using an existing supported interface do not need a CLI update. A new protocol or result format can require one.

Find models

grida ai models list --modality image
grida ai models list --provider fal --modality video --local-image

Rows identify provider, model, input variant, status, and accepted local-image flags. --local-image selects operations accepting a local --reference or --image file. It does not mean every model from that provider accepts files. Use --json for structured rows and --kind for operations such as music, sound-effect, text-to-speech, or three-d.

Check your access route

grida ai models list --provider fal --available
grida ai models list --provider openrouter --modality image --available --json

--available requires a provider. For BYOK it checks effective key presence after static validation, without contacting the provider. For GG it checks cached organization eligibility online using your Grida session. Neither guarantees provider permission, provider credit, or a successful request. A missing provider key does not prevent offline discovery or use of another configured provider.

Inspect accepted inputs

grida ai models inspect --provider openrouter --model openai/gpt-image-2 --variant references
grida ai models inspect --provider fal --model google/veo-3.1-lite --variant image
grida ai models inspect --provider fal --model google/veo-3.1-lite --variant image --json

Human output shows types, required fields, limits, and an example. --json returns the full schema. GG JSON operations also include the current published usage tariff and its units, rather than an installed provider-cost estimate. The selected --model supplies the canonical request's model field; do not repeat it in the input JSON. Inspect the intended variant: image/video inspection defaults to text input where that variant exists. Generation's --reference and --image flags select compatible variants automatically.

GG's generic adapter supports JSON image generation, video, music and 3D operations. Multipart-only image editing is not advertised by that adapter; existing native Tripo file and rigging operations retain their separate contracts. GG discovery does not require a separate Gateway API key. Generation uses your Grida session and selected organization.

The selected operation's schema is authoritative. A provider's model page may advertise options Grida does not expose. There is no raw provider passthrough or automatic provider fallback.

Find speech voices

grida ai voices list --provider elevenlabs

Voice discovery uses your ElevenLabs key and needs a connection. Pass a returned ID through --voice when generating speech.