> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rimp.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Changelog

> API changes, new models, and SDK releases. Follows Keep a Changelog.

## 2026-05-22 · 1.1.1 — Status page, retries, and a pricing refresh

### Added

* **Live status page** at [status.rimp.io](https://status.rimp.io) — a real-time snapshot of API and Studio health, refreshed continuously.
* **Try again on failed generations** — async generations that error out now expose a one-click retry. Retries reconcile with the provider first so you're not double-billed.

### Changed

* **Catalog pricing refreshed** — base credit costs for 16 models updated to match the latest underlying provider rates. Kling v3 repricing included. Call [`GET /v1/models`](/api-reference/list-available-models) for the current numbers.
* **Plan margins repriced** — Pro and Studio now use a ×2.5 margin (up from ×1.6); the Studio plan moves from $100 to $120/month. Free and Team are unchanged. See [Multi-model](/concepts/multi-model) for how margins apply.
* **Avatar catalog tightened** — HeyGen avatars are now filtered to Avatar V-compatible looks, so every avatar listed actually renders.
* **3D outputs default to GLB** — `POST /v1/three-d` returns the final GLB mesh directly instead of a color preview.
* **Upscale storage** — upscale outputs are re-encoded PNG → JPEG (q90), cutting download size 5–10× with no visible quality loss.

### Removed

* **14 unavailable models delisted** — models that no longer run reliably upstream were removed from the catalog after a live audit. They no longer appear in [`GET /v1/models`](/api-reference/list-available-models) and will return `model_unavailable` if referenced.

### Fixed

* Community model versions are now pinned, fixing intermittent `404` errors on generation.
* Avatar adapter input shapes corrected — avatar jobs no longer fail validation against the provider schema.
* Output `type` is now derived from MIME when the upstream API omits it, so downloads always have the right extension.
* Chat streaming stays alive past the initial response and no longer truncates long completions.
* Server-Sent Events for in-flight generations are more reliable: no more reconnect loops on transient failures, and failed cards are replaced cleanly on retry.
* Studio grid hydrates from the server when you switch modalities, so you always see your latest generations.

## 2026-05-20 · 1.1.0 — Six new modalities

### Added

* `POST /v1/voice` (sync) — text-to-speech and voice cloning.
* `POST /v1/music` (async) — full songs, with or without vocals.
* `POST /v1/avatars` (async) — talking avatars and lipsync.
* `POST /v1/chat` (sync) — LLM chat completions across 15 models.
* `POST /v1/three-d` (async) — text-to-3D and image-to-3D, GLB mesh output.
* `POST /v1/upscale` (async) — image upscaling (2x / 4x) and face restoration.
* `GET /v1/comparisons/{id}` — fetch a comparison with all child generations.
* Auto-generated [MCP server](/mcp) — point AI tools at the docs.

### Models

* Catalog expanded to \~60 models across **8 modalities** (image, video, voice, music, avatar, chat, 3D, upscale). See [Multi-model](/concepts/multi-model) for the full tables.
* New image models include Flux 2 (Pro / Max / Klein), Imagen 4 (+ Ultra), Nano Banana Pro, Recraft V4, Qwen Image 2, Seedream 5 and gpt-image-2.
* New video models include Sora 2 (+ Pro), Kling v3 (+ Omni), Seedance 2.0 and Runway Gen-4.5.

### Changed

* Plans are now **Free / Pro / Studio / Team**. Comparison caps: 2 / 3 / 6 / 12 models.

## 2026-05-19 · 1.0.0 — Initial public API

### Added

* `POST /v1/images` (sync) — single-model image generation, returns 200 with outputs inline.
* `POST /v1/videos` (async) — single-model video generation, returns 202 with `generation_id`.
* `POST /v1/comparisons` (async) — multi-model side-by-side, atomic credit reservation.
* `GET /v1/generations/{id}` — poll status + signed output URLs.
* `POST /v1/generations/{id}/cancel` — cancel queued or running generations.
* `GET /v1/models` — live catalog with pricing and capabilities.
* `POST /v1/uploads` — presigned PUT URLs to Cloudflare R2 for image-to-image / image-to-video inputs.
* `GET|POST|DELETE /v1/webhooks` — subscribe to `generation.completed`, `generation.failed`, `comparison.completed`, `usage.budget_threshold`, `usage.anomaly_detected`.
* `GET /v1/usage` — PAYG usage aggregated by day / model / api\_key.
* `Idempotency-Key` header on all credit-consuming POSTs (24h replay window).
* HMAC-SHA-256 signed webhooks with `X-Rimp-Signature: t=…,v1=…`.

### Models

* **Image**: `openai-image`, `stability-sd3`, `flux-pro`, `flux-pro-i2i`, `ideogram-v2`, `ideogram-v2-turbo`.
* **Video**: `veo-3-1-fast`, `veo-3-1-standard`, `runway-gen4-turbo`, `runway-gen4`, `luma-ray2`.

### SDKs

* `@rimp/sdk` for TypeScript / JavaScript — covers all endpoints with full types.
* `rimp` for Python (3.9+) — sync client over httpx.
* curl recipes published in [SDKs / curl](/sdks/curl).

### Enterprise

* WorkOS SSO (SAML / OIDC) + SCIM provisioning.
* Append-only audit log with CSV export.
* IP allowlist per API key.
* BYOK — bring your own OpenAI / Stability / Veo / Runway / Luma keys.
* Custom Stripe invoicing with PO numbers.
* SLA 99.9% with automatic credit accruals.