NNexithDocs
StatusDashboard ↗
Overview & quickstartModel catalogGuidesAPI referenceIntegrationsOAuth2 / OIDC

Nexith Image

Text-to-image and image editing Model ID: nexith-image

Performance Speed Price
High quality About 40–90 seconds/image $0.10 / image
Input text, images Output images Size 1024×1024

nexith-image generates images from text prompts and also accepts a reference image for editing (img2img). It is not a synchronous API: submission immediately returns 202 with a job ID, which a worker executes asynchronously; clients poll the job or subscribe to events over SSE to get artifact URLs.


Pricing#

Item Price
Generation $0.10 / image
Editing $0.10 / image
  • Submitting a job reserves balance at the unit price; it is settled on completion, and failures/cancellations are fully refunded.
  • Insufficient balance returns 402 insufficient_balance before enqueueing and never reaches the upstream.
  • Artifacts are stored in object storage (S3) and returned as presigned URLs valid for 7 days; copy them yourself if you need long-term storage.

Modalities#

Modality Support
Text Input only (the prompt)
Images Input (editing reference) + output
Audio / video Not supported

Endpoints#

Endpoint Description
POST /v1/images/generations Submit a generation job
POST /v1/images/edits Submit an editing job (chat site JWT only; nx- keys get 403)
GET /v1/jobs/:id Query status and artifacts
GET /v1/jobs/:id/events Event SSE
POST /v1/jobs/:id/cancel Cancel

See the Image guide for the full flow and examples.


Capabilities#

Capability Supported Notes
Text-to-image ✅ Any text prompt
Editing (img2img) ✅ Upload a reference image → nx-file://<id> → /v1/images/edits
Streaming ❌ Async job model
Function calling ❌ —
Multi-image output ❌ One image per job

Rate limiting#

Uses the same (API key, model) rate limiting chain as text models; exceeding it returns 429 before enqueueing, without locking funds or creating a job row. The concurrency ceiling is determined jointly by the worker and the upstream channel; at peak times jobs queue in pending, and you can observe status changes via GET /v1/jobs/:id.