VideoGenReview

What each AI video model's release notes say changed

Changelog terms: the words release notes use

Each page defines one term as the notes use it and points to entries that show it. As of 2026-09-25.

Inside this sectionThe terms AI video release notes use, preview, snapshot, turbo, keyframe, and what they mean there. The section holds 16 pages.SectionChangelog terms16 pagesEach pageOne term as the notes use it.Each entryNote address, note date, day read.
Fig. 1 Every page in the section links back to this one.

Definitions follow the makers' own usage where it is consistent.

Where makers use a term differently, the page says so.

1In this section

  • Audio-to-video — Audio-to-video generates a clip driven by an audio track, so speech or music shapes what happens on screen.
  • Batch and async — Batch APIs accept many generation requests at once and return results later; async endpoints return a job to check on instead of waiting for the video.
  • Director models — A director model is trained to follow camera instructions written in the prompt, rather than taking them as a separate setting.
  • Extension — Extension continues a finished clip with a further generation that picks up where the clip ended, so it grows in steps.
  • General availability — General availability, often shortened to GA, marks a model as stable and supported for production use, after a preview.
  • Keyframe — A keyframe is an image the generated video must pass through: the first frame, the last frame, or a frame in between.
  • Limited public beta — A limited public beta opens a model to a restricted group of outside users before a wider release, so their feedback can shape it.
  • MCP server — An MCP server exposes a service's tools through the Model Context Protocol, so an AI agent or chat client can call them directly.
  • Multi-shot — Multi-shot, as AI video release notes use it: multi-shot output holds several shots, with cuts between them, in one generated video or one request.
  • Native and upscaled resolution — Native resolution is the size the model renders at; upscaled resolution is a larger size made from a smaller render afterwards.
  • Preview — Preview, as AI video release notes use it: a preview is a release the maker marks as early: available to use, but with the maker keeping room to change it.
  • Reference-to-video — Generating a video from one or more reference images or clips that fix how a person, object or place looks, with a prompt for what happens.
  • Snapshot — Snapshot, as AI video release notes use it: a snapshot is a dated, fixed build of a model.
  • Standard and pro modes — Two quality settings on one model, a quicker standard mode and a slower, higher-quality pro mode, often with features arriving in one mode first.
  • Turbo, Fast, Flash, Lite, Mini — Faster or cheaper versions of a model released alongside or after it, under names such as Turbo, Fast, Flash, Lite, Mini or Prime.
  • Video to video — Video to video takes an existing clip as input and changes its look or content while keeping its movement.

2Other sections