SpiderGate adds image, video and speech generation to its LLM gateway

WILMINGTON, USA, July 13, 2026 — SpiderIQ today announced that it has released media generation through SpiderGate, its multi-provider LLM gateway, so that images, video, speech, transcription and embeddings run on the same API token, billing record and trace view customers already use for chat, detailed at https://gate.spideriq.ai/media-api. Thirty-three of the eighty-one media models registered in the SpiderGate catalogue are active and callable at launch.

The addition targets a specific operational cost. Teams that route text through a gateway have typically kept a separate account, a separate key and a separate invoice for image or audio generation, because those providers expose different request shapes. SpiderGate now accepts both a passthrough call compatible with the OpenAI SDK and a schema-aware call that reads each model's own declared parameters.

The schema is the part aimed at autonomous agents. A request to the model-listing endpoint returns every active model together with the exact parameters it accepts, including types, defaults and permitted values. An agent reads that contract and sends only those fields.

Two validation rules run before a provider key is selected, according to the published API reference at https://gate.spideriq.ai/docs/reference/api. A parameter a model does not declare is dropped before the provider is contacted and named back to the caller. A declared parameter carrying an invalid value is rejected with a typed error that names the field and lists what it will accept. Because both checks precede key selection, a malformed request does not consume a paid generation.

"An agent should not have to guess what a model accepts, and it certainly should not find out by spending a generation," said Martin Shein, founder of SpiderIQ. "We published the parameter schema as data, so the controls a person sees in the dashboard and the fields an agent sends over HTTP come from the same row in the same table. Nobody maintains a second copy."

Coverage is stated as a fraction rather than a total. Of the 81 media models registered in the catalogue, 33 are active: 16 of 34 image models, 15 of 36 video models and 2 of 11 audio models. The remaining 48 are registered with a recorded reason, such as a provider-side error or a generation slower than the platform's polling ceiling, and are not yet servable.

Media generation is bring-your-own-key. Every active media model is paid tier and none draw on SpiderGate's shared key pool, so customers register their own credentials for the third-party providers they wish to use, which currently include kie.ai, fal.ai, Google, xAI and OpenAI. SpiderGate supplies the routing, the parameter schema, the validation, the timeout guards, the usage metering and the storage of the generated asset.

The release also extends SpiderGate's task-alias system to embeddings. Three aliases resolve on the server to the corresponding embedding model, so retrieval code no longer has to name a vendor directly. Existing calls that pass a model name continue to work unchanged.

"Every illustration in our announcement post was generated through this API, which is the test I care about," said Martin Shein, founder of SpiderIQ. "If we would not use it for our own work, it is not ready for anyone else's."

Generation runs under a wall-clock limit, documented at https://gate.spideriq.ai/docs/using-the-gateway/media-generation, of 600 seconds for image and speech and 900 seconds for video, after which a request returns a typed timeout rather than hanging. Each dispatch failure is recorded with an error type, so a model that has begun failing is visible as a number rather than as silence.

The Media API is available now to SpiderGate customers.

About SpiderIQ

SpiderIQ builds infrastructure for AI agents, including SpiderGate, a multi-provider LLM gateway that handles routing, key management, budgets and observability across a catalogue of more than one thousand models.

Media contact

Press Office, SpiderIQ — press@spideriq.ai — https://gate.spideriq.ai/press

###