AVE 1.0.25 Release Notes

Status: Released on August 3, 2026 Bangkok time.

Highlights

Added media generation with Google Veo, Google Nano Banana, Runway, Luma, Kling, fal.ai, and Seedance through capability-driven provider controls.
Added text-to-video, image-to-video, image generation, and image editing where supported, with provider-specific models, aspect ratios, durations, resolutions, reference inputs, audio options, and advanced controls.
Added in-editor access to the global Asset Library with semantic search, asset previews, and one-click import into the active project.
Refined the editor's primary actions, header hierarchy, Assistant panel controls, asset scrolling, timeline lane alignment, separators, and toolbar interaction states.
Improved direct Ask-mode audio editing so trim, extend, and fade actions can execute immediately when the request is complete.

Base range: 1c727f6..3fe5b88.
Advanced the complete AVE package from version 1.0.24 build 26 to version 1.0.25 build 27.
Added secure bring-your-own credential fields and provider-specific generation adapters for Google Veo, Google Nano Banana, Runway, Luma, Kling, fal.ai, and Seedance.
Kept the existing Local AI runtime, starter model payload, Sparkle signing identity, minimum update build 18, and full-package update path unchanged.
This remains a complete AVE package distributed through lemberalla/ave-releases-v2; the legacy compatibility bridge remains untouched.

Release verification includes:

Full renderer and render-worker builds, typecheck, AI prompt checks, regression suite, dependency audits, package audit, signing, notarization, stapling, DMG integrity, and Sparkle signature verification.
Provider-capability validation across generation modes, model selection, parameter normalization, secure credential handling, job lifecycle, cancellation, output download, and asset import.
In-project Asset Library search, preview, semantic result loading, project import, and stable card/layout behavior.
Direct Ask-mode audio trim, extend, and fade actions with immediate timeline and inspector synchronization.
Installing the exact release DMG before publication and testing the update from the previously published full AVE build immediately after publication.

External generation requires the user's own credentials and is subject to the selected provider's availability, account limits, pricing, content rules, and processing times.
Generation modes, models, resolutions, durations, reference inputs, and audio options vary by provider; AVE adapts the form to the selected capability.
Generated outputs are imported as normal AVE assets so they can be searched, arranged, edited, and exported alongside other footage.
The complete download remains large because it includes AVE's Local AI runtime and starter models.