SDK 3.0.0 — face-ref merged into character-ref, reference tools on the Gen8 image engine, caps 6/6/6 and 6 subject images
2.8.3 — Uploads: .m4v clips (stored as mp4), PROTECTED_MEDIA / INVALID_MEDIA documented (Uploads, Home, Errors)
Gen8 Preview images: priced on render time measured on the standard workers, roughly four to seven times the Gen 7.2 preset (prices of 2026-10-12)
Gen8 Preview images: priced on render time, roughly three times the Gen 7.2 preset (prices retuned 2026-10-11)
negativePrompt: applied by Gen8 Preview images since 2026-10-11, the Gen7 image engine renders without it
Gen8 Preview images (SDK 2.8.0): imageModel next, 6 references, startedAt, negativePrompt note, Gen8 Preview naming
2.7.0 — audio reference for Gen8 Flash: uploads.audio, generate.video audioCreationId / audioUse, mediaKind and duration on creation rows, REST fields and errors, derived idempotency token
2.6.1 — enhanced on generate.video (raw style), video styles vs TIXI rule, 'tixi ' thinking pass
API reference: videoModel lives on the queue state, a Gen8 Flash creation is stamped aiModel gen8.0
2.6.0 — Gen8 Flash (Preview): videoModel 'next', account-driven lengths and presets, REST fields and errors, recipe; prompt caps up to 20,000; video lengths 5/6/7 s
Content rating: a flagged result is kept, masked and unpublishable — not an error
Remove the internal staging note page
2.5.6 — errorCode + errorMessage on failed jobs; waiting for render capacity Failed, timed-out and expired jobs now carry errorCode (to branch on) and errorMessage (to show), on QueueState and GenerationFailedError, with the code table in Errors. Generating, REST-API and Troubleshooting describe how a job waits for render capacity and when it ends expired (refunded).
Video auto draws max / ultra only — never ultimate
1.4.6 / 2.5.5 — batch check_generation, full get_creation, real clip lengths, price formula, resolved quality on submit, CREATION_NOT_FOUND
Recommended presets (max / ultra 5 s / ultimate), 2,500-char prompts, 5 s video default, account presets in get_account, agent skill
REST API: quality 'auto' draws max / ultra / ultimate by tier
REST API: quality 'auto' draws max / ultra by tier, never ultimate
groupId lens — one listing or a random pick across a folder group
Artist Painter: artistic styles (list + generate), unsigned by default, 24 tools
Community links in the sidebar
Gen 7.2 — model references updated
v2.3.2 — About the models (Gen 7.1, 2DAI❤️ART, private cloud)
v2.3.0 — listing engine, creation actions, folder groups, stats + read-only finance
v2.2.0 — generate.similar, creations.clone, usedRef filter, queue archive transparency
v2.1.0 — integration option, CDN resize, folders pagination, User-Agent contract, corrected error statuses
Wiki: REST API reference — endpoints, payloads, error codes and limits
Wiki: verified HTTP statuses + exact scope/spend-cap denial payloads
Wiki v2.0.1 — verified error paths, reference lifetimes, source field, testing recipes
Wiki v2.0.0 — new SDK documentation