Skip to content

Wan 2.5

model version · in 14 sources (2025–2026)

Wan 2.5 is a video model on Higgsfield that makes clips of up to ten seconds at 480p, 720p or 1080p, from text or an image reference, with synced audio and lip motion. It arrived on Higgsfield in September 2025; in December 2025 tests Adil found it decent but niche, prone to artifacts and weak on consistency, and overpriced once Kling 2.6 was out.

Part of: Wan · Includes: Wan Camera Control

Sources by year: 2025: 5 · 2026: 9

Threads

Also named in the threads: Kling on Higgsfield · Lipsync Studio on Higgsfield

Often together

What comes up in the same parts of the videos; the number is how many parts.

Veo 3 (5) · Kling 2.6 (5) · Sora 2 (5) · LipSync Studio (5) · lip sync (6) · text-to-video (3) · Google (4) · native audio (5) · Higgsfield Audio (3) · Reframe (3)

What the videos say

The most important statements of each year; all statements.

2026

  • 2026-07-07 · How to Make Realistic AI Talking & LipSync Videos in 2026 (Avatars, Voice & Consistency) — recommendation: Higgsfield: For multilingual talking videos, Higgsfield's blog notes HeyGen's 175+ languages and LipSync Studio's Wan 2.5 Speak and Kling Avatars 2.0, and advises testing on the target language before a full run, since quality varies by language.
  • 2026-08-01 · How do I use lipsync, voiceover, and aspect ratios? — Higgsfield: Lipsync Studio models as of 1 August 2026: image-to-video Kling 2.6 Lipsync (up to 1080p, with audio), Wan 2.5 Speak (480p to 1080p), Kling Avatars 2.0 (longer clips), Higgsfield Speak 2.0, InfiniteTalk (long-form); video-to-video Kling Lipsync and Sync Lipsync 3 (up to 4K).
    • Higgsfield lists FLUX 3, MiniMax Hailuo 3.0 and Wan 2.5 as models that generate native audio with the video; Wan 2.5 syncs sound with camera movements.
  • 2026-08-24 · Changelog: Wan 3.0 and Wan 3.0 Prime. Now on Higgsfield — announcement: Higgsfield's changelog of 24 August 2026 says Wan 3.0 and Wan 3.0 Prime, the newest Wan video models, are live on Higgsfield, generating video from text and images.
  • 2026-09-18 · How to Create 100+ Ad Variations With AI: Swap Hooks, Products, and Languages — Higgsfield's blog says Lipsync Studio supports 18 languages across seven models (Veo 3, Wan 2.5 Speak, Kling Avatars 2.0, Higgsfield Speak 2.0, InfiniteTalk, Sync Lipsync 3, Kling Lipsync); Veo 3 and Wan 2.5 Speak each add 15+ presets for formats and camera setups.
  • 2026-10-08 · How to Take a Character From Still Image to Talking Video Without Losing Their Identity — Higgsfield's blog prices a 10-second 1080p talking video of one character: $1.25 one-time character setup, 1.2 credits ($0.06) for a Seed Audio 1.0 line, $2.00 for the Wan 2.5 Speak video; the first video 66.2 credits ($3.31), each later one 41.2 ($2.06).
    • For a Wan 2.5 Speak talking video in Lipsync Studio, Higgsfield's blog advises starting with the General preset at 1080p and 10 seconds.
  • 2026-10-22 · introducing Wan 3.0 — Higgsfield: Wan 3.0 builds up to six linked shots with sound from one prompt in a single pass, keeping the same face, light and location throughout.
  • 2026-10-22 · wan-2.6 — Higgsfield: Given a description of a full scene, Wan 2.6 handles the camera transitions and generates a 15-second video (as of 2026-10-22).
  • 2026-10-22 · WAN 2.5 × Higgsfield — Higgsfield: Wan 2.5 output on Higgsfield (as of 2026-10-22) comes synced with audio, includes lip motion, and is offered at 480p, 720p and 1080p.
    • Higgsfield says Wan 2.5 supports clips of up to ten seconds in multiple aspect ratios, beyond the eight-second cap of Google Veo 3, while being faster and more affordable.
    • Higgsfield's Wan 2.5 page (as of 2026-10-22) says the model creates HD videos with synchronized voice, audio and real camera movement in seconds.
  • …and 5 statements more in 2026 — all statements of 2026

2025

  • 2025-09-26 · Meet Higgsfield x Wan 2.5 — announcement: Higgsfield introduces Wan 2.5 on Higgsfield. — video title
  • 2025-10-11 · Higgsfield x Wan 2.5 Camera Control — announcement: Higgsfield introduces Wan Camera Control and says it is Unlimited. — video description
  • 2025-10-13 · Master Sora 2 for AI Commercials And Grow Audience — Ruslan: Video models on Higgsfield as of 2025-10-13 include its own camera-preset models, Sora 2, Kling, Wan 2.5, Veo and MiniMax, each with many presets. — 00:06:18
    • Ruslan: For now, to put your own face in a video, Ruslan suggests Veo 3 or Wan 2.5 (as heard), whose clips are less dynamic than Sora 2's. — 01:06:01
  • 2025-12-14 · WAN 2.6 by Alibaba is Here: A User Guide to AI Video Model — announcement: Higgsfield: Alibaba's Wan 2.6 is in Higgsfield's video model list as of 2025-12-14.
  • 2025-12-16 · Kling 2.6: This Cheap Model DESTROYS Veo 3.1 — Adil: In the camera-control tests Veo 3.1 got confused and put the drone inside the shot, and Wan 2.5 struggled with consistency. — 00:02:50
    • In the shadowboxing test, Wan 2.5 produced artifacts, with the background and the gloves warping. — 00:03:58
    • For animating humans or precise body mechanics, Adil recommends Kling 2.6 as smoother and more consistent than Veo 3.1, Wan 2.5 and Sora 2. — 00:06:02
    • In the physics tests, Veo 3.1 and Wan 2.5 were capable but struggled to keep things stable. — 00:06:02
    • Adil finds Wan 2.5 decent but niche, overpriced now that Kling 2.6 is out, with weaker sound. — 00:08:58
  • …and 3 statements more in 2025 — all statements of 2025