Feature Request: Add OpenArt as Generation Provider

[kyanitelabs]

Summary

Add OpenArt as a generation provider alongside Seedream, Seedance, GPT-image-2, Gemini 3 Flash Image, Imagen, Veo 3, Kling, and OpenRouter.

Why

OpenArt is an all-in-one AI Creator Studio that aggregates multiple top-tier models (Google Veo, Kling, Seedance, Nano Banana Pro, GPT-image-2, and more) behind a single API key. This aligns directly with pushing-frames' bring-your-own-keys philosophy — one key, many models.

Benefits for pushing-frames users

• Economical access to premium models — Users who already pay for OpenArt subscriptions can use their existing accounts instead of purchasing separate API keys for each individual provider. This is a real cost concern when running multi-shot cinematic packs.
• Multi-model rapid prototyping — OpenArt lets you hit several model aesthetics through one interface. A director can test whether a concept works across different model looks before committing to final shot generation.
• Character & world consistency — OpenArt has built-in character and world consistency features (persistent visual identity across generations). This is relevant for cinematic packs where the same subject appears across multiple shots — currently a pain point when each provider handles identity differently.
• Image + video + audio — OpenArt covers all three media types, which maps well to the growing scope of cinematic packs (storyboard stills → final video → potential audio/soundtrack integration).

How it could work

Following the existing BYOK vault pattern:

1. User adds their OpenArt API key to the vault (encrypted, stored locally in IndexedDB, never leaves the machine — same as existing providers)
2. OpenArt appears as a provider option in the shot generation UI
3. Under the hood, OpenArt's multi-model routing means the user can optionally specify which underlying model OpenArt should use per-shot, or let OpenArt's defaults handle it
4. Generated assets land in generations/ as usual

Provider adapter scope

A new provider adapter at src/providers/openart.ts (or similar) implementing the existing provider interface:

• API auth via key from vault
• Model selection (OpenArt exposes: Veo, Kling, Seedance, Nano Banana Pro, GPT-image-2, Imagen, etc.)
• Image and video generation endpoints
• Cost estimation (pre-flight cost checks, respecting project budget caps)
• Reference image conformance (resize, byte-cap, format per OpenArt's requirements)

OpenArt API reference

Docs: https://openart.ai — API access available to account holders.

---

I have an active OpenArt account and would use this as my primary generation provider. Happy to help test an adapter if this gets picked up.