Launch
- improvedAlibaba Cloud Model Studio is now the native engine for image and video generation — one provider, predictable cost, no surprise billing
- newHappyHorse 1.0 flagship video — ranked #1 on Artificial Analysis Video Arena, supports text-to-video, image-to-video, and reference-to-video
- newZ-Image-Turbo for fast standard image generation — top open-source benchmark, 8-step inference, ~5 seconds per image
- newNative synced-audio video via wan2.7-i2v — generate a clip with built-in narration in one call (no separate voiceover + stitch step)
- newCost-accurate per-call credit deduction — every generation deducts credits matched to the actual provider cost × margin, not a flat rate
- newConversational BYOK picker — when you have your own Higgsfield / Replicate / Vertex key connected, Copilot asks once which to use and remembers for the rest of the conversation
- improvedSettings → Integrations reorganized into Native engines vs BYOK-only providers, with honest copy about what works without each key
- fixedKilled all silent provider fallbacks — generation never silently escalates to a more expensive provider when the cheap one is unavailable
- fixedEntity cover + canvas cover + standalone image generation all routed through the same native engine