Skip to content
Bud's Wiki

Wan 2.7 upcoming features preview (video editing, V2V, real-time)

TL;DR

Section titled “TL;DR”

A short “wan 2.7” preview tweet from @bdsqlsz (青龍聖者) — a reliable community leaker for the Wan/Alibaba video stack — showing an image with the upcoming feature roster for Wan 2.7. Per third-party API listings Wan 2.7 supports advanced features like first-and-last frame control, 3x3 grid synthesis, and instruction-based video editing, alongside first-and-last-frame generation, video-to-video editing, and subject referencing, plus support for real-person image inputs, up to five video references, and 1080P high-definition outputs spanning 2 to 15 seconds. Paul flags it as non-research but a useful early signal on where the open-weights Wan family is going.

Key claims

Section titled “Key claims”
  • Tweet is a single-image teaser of “wan 2.7” upcoming features [tweet body].
  • Confirmed in downstream API listings: V2V (video-to-video editing), first/last-frame control, 9-grid (3×3) image-to-video, instruction-based editing, up to 5 video references, 1080p, 2–15s clips [external context — atlascloud, wan2-7.io].
  • Author @bdsqlsz is a known reliable community source for upcoming Wan/Qwen-stack releases (per @Paul) [pointer note].

Method

Section titled “Method”

Product preview — no methodology. The tweet itself contains the string “wan 2.7” and an attached image listing features. The substantive content lives in the image and in the downstream API documentation that followed.

Results

Section titled “Results”

Not applicable — this is a leak/preview, not a paper or release with measured results. Pointer value is “what to expect from the next Wan version”.

Why it’s interesting

Section titled “Why it’s interesting”

Wan 2.7 is the natural successor to the Wan-2.x family that Luma’s video research already tracks via Wan-Animate: Unified Character Animation and Replacement with Holistic Replication (Wan-Animate, character animation/replacement on Wan-2.x). The advertised feature set — V2V editing, multi-reference conditioning, first/last-frame control, instruction-based edits — sits at the intersection of World Foundation Models (Wan as an open-weights video FM) and Layered Image/Video Decomposition/Camera-Controlled Video Diffusion style controllability. The “real-time” angle Paul highlights also connects to the recent push toward Helios: Real Real-Time Long Video Generation Model and Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation-style few-step interactive video generation — worth watching whether Wan 2.7 ships a distilled variant in that regime.

See also

Section titled “See also”