Seedance 2.5 has arrived in 1080p on Higgsfield, CapCut and Dreamina, and free generations are live right now for new users
The model that already made headlines for 30-second single-shot generation has just stepped up to full 1080p resolution, and three major platforms went live with it simultaneously on 14 August 2026.
The resolution ceiling on Seedance 2.5 just came down. On 14 August 2026, three platforms simultaneously announced 1080p access to the model that already turned heads for generating full 30-second shots in a single pass: Higgsfield, CapCut, and Dreamina. That is not a staggered rollout. That is a coordinated upgrade, and for anyone generating commercial or short-form content with AI, it matters more than another benchmark score.
The jump from 720p to 1080p sounds incremental on paper. In practice it is the difference between footage you have to apologise for in a client deck and footage you can drop straight into an edit. Seedance 2.5 already had the temporal consistency to hold a 30-second shot together. At 1080p, individual frames can now withstand a pause, a zoom, or a broadcast compression pass without falling apart. That is the threshold most commercial workflows require before they will even consider AI-generated material.
What each platform is offering
@higgsfield_ai is positioning itself as the exclusive early-access home for the 1080p upgrade, and is backing that claim with free 1080p generations for all new users at zero credit cost. The offer is described as limited time with no end date publicly stated. Higgsfield's demos show full 30-second commercials, from food advertising close-ups to cinematic product spots, generated from a single text prompt. The platform has also published a companion sci-fi short, Adiliada, made entirely on Cinema Studio 4 using the same model.
@capcutapp has rolled out Seedance 2.5 in 1080p across its app, web, and desktop surfaces simultaneously, covering AI Lab, AI Generator, AI Video, AI Edit, AI Extend, and Video Studio. Geographic availability is currently limited to Asia, the Middle East, Africa, Europe, and South America for subscribers. North America is not yet listed, which is a notable gap for the platform's largest creator market.
@dreamina_ai has launched the same 1080p tier on Dreamina Web, also restricted to Southeast Asia, the Middle East, Africa, Europe, and South America at launch, with more regions described as coming soon. Dreamina is also running a Seedance 2.5 3D Reference Creative Contest with a jury that includes Emmy Award-winning director Paul Trillo and director Lauren Indovina, whose work is in MoMA's permanent collection.
What this changes for a real workflow
The short version: the footage you generate today is more likely to cut with footage you shot on a camera. That has not been reliably true of AI video at any length above a few seconds. At 30 seconds with consistent characters and 1080p output, the gap between AI-generated b-roll and traditionally shot material narrows to a point where editorial decisions become creative rather than remedial.
For commercial directors specifically, Higgsfield's framing around food advertising is pointed. Food commercials are among the most lighting-sensitive categories in commercial production, where reshoots are expensive and shot windows are narrow. If the model can hold specular detail on a liquid pour or a food texture across a full 30-second take at 1080p, that is a genuine production cost story, not just a demo.
What was not said
No pricing structure for 1080p generations has been announced beyond the new-user free tier on Higgsfield. It is not yet clear what credit cost 1080p will carry once the introductory offer ends, nor whether the free tier applies to existing users or only new sign-ups. CapCut and Dreamina have not specified per-generation pricing for the 1080p tier.
Geographic availability on CapCut and Dreamina excludes North America at launch, with no date given for expansion. Users in the US, Canada, and Australia cannot access the 1080p tier on those platforms right now.
No technical specifications have been published about what changed internally to reach 1080p output. Whether this is a higher-resolution decode from the same latent, a separately trained checkpoint, or a post-processing upscale is unknown. That distinction matters for quality, but there is no public documentation to answer it yet.
What to test first
If you are a new user on Higgsfield, the zero-credit offer makes the answer obvious: run your most demanding commercial brief, ideally something with product close-ups, fluid motion, or consistent branding across cuts. That is where 1080p resolution either earns its place or exposes the model's limits. Pause on individual frames, inspect texture and edge detail, and compare directly to whatever you were generating at lower resolution last week. The Higgsfield food and drink demos are the implicit benchmark the platform is setting for itself, so hold it to that standard.