The idea is to cover three common situations: fast, cheap idea exploration (Seedance 2.0 Mini), quick previews inside the Kling family (Kling 3.0 Turbo), and people-in-frame video with synchronized sound (HappyHorse 1.1).
Kling 3.0 Turbo: a fast tier for drafts
Kling 3.0 is Kuaishou's third-generation video line: stronger frame consistency, more photorealistic output, and clips up to 15 seconds. Turbo is the fast tier of that line — run quick takes to test motion and framing, then rerun the best idea on a higher Kling tier.
Text-to-video and image-to-video
Generate from a description or animate a starting frame.
3–15 seconds, 720p/1080p
Any length in that range; 16:9, 9:16, and 1:1.
Built for speed
Faster than higher Kling tiers — for quick drafts before a final render.
Seedance 2.0 Mini: cheap and fast, for volume
Seedance 2.0 is ByteDance's multimodal video model — text, image, video, and audio in; multi-shot clips up to 15 seconds with native sound out. It currently ranks first for video with sound on the independent Artificial Analysis arena, in both text-to-video and image-to-video. Mini is its most affordable tier: built for volume and speed — fast drafts, concepts, and short-form — with the option to rerun a favorite on a higher Seedance tier for final quality.
Text, image, and video reference
Generate, animate a frame, or guide motion with a video reference; up to 9 image references.
Native sound, toggleable
Sound is generated with the video; in Mini you can turn it on or off.
480p/720p, 4–15 seconds
Light resolutions tuned for speed and volume.
HappyHorse 1.1: people in frame and native sound
HappyHorse is an Alibaba video model that appeared anonymously on the Artificial Analysis arena in spring 2026 and took first place for visual quality in the no-audio ranking; Alibaba later confirmed authorship. HappyHorse 1.1 is the updated version. Its strength is people in frame — faces, expressions, speech — and it generates sound natively, together with the video in one pass. With sound, 1.1 sits right behind Seedance 2.0 near the top of the arena.
Text, image, and references
Generate from a description, animate a frame, or work from a set of visual references.
Up to 9 references
Lock a character, outfit, and style with reference images held across the clip.
Native synchronized sound
Audio is generated with the video — built into the architecture, not a separate step.
An updated Alibaba model focused on people in frame, with native synchronized sound. Up to 9 references for a consistent character, 720p/1080p, up to 15 seconds.
Topaz Astra: enhance or restyle a finished clip
Topaz Astra solves a different problem — enhancement and restyle, not generation. Astra 2 takes a finished clip and returns an improved 1080p or 4K version. An optional prompt can describe the intended restyle; creativity, sharpness, and realism tune the result. It is available in Givon AI as a video-to-video tool.
Enhance or restyle at 1080p/4K
Choose 1080p or 4K output while improving detail and texture in a finished clip.
Removes the plastic AI look
Adds believable texture so frames look less synthetic.
Video → video, optional prompt
Upload a finished clip; add a prompt when you want to describe a restyle or enhancement.
Creativity, sharpness, and realism
Tune how strongly Astra changes the source and how much detail and natural texture it preserves.
Which model to use
Explore many ideas cheaply — Seedance 2.0 Mini
The most affordable tier: fast drafts, concepts, and short clips.
A quick Kling-style preview — Kling 3.0 Turbo
Test motion and framing fast, then rerun the best take on a higher Kling tier.
People in frame and sound — HappyHorse 1.1
Faces, expressions, speech, and native synchronized sound in one pass.
Enhance or restyle a finished clip — Topaz Astra
A video enhancement and restyle pass with optional prompt, 1080p/4K output, and creativity, sharpness, and realism controls.
You don't have to pick the model up front: run the same scene through several and keep the take that holds style, motion, and composition best.
Questions about the new models
Are these new models or just variants?
Both. Kling 3.0 Turbo, Seedance 2.0 Mini, and HappyHorse 1.1 are new tiers and versions of familiar families — faster, cheaper, and newer. Topaz Astra is a genuinely new tool: it improves finished video rather than generating it.
Which of them have sound?
HappyHorse 1.1 always generates native sound. Seedance 2.0 Mini lets you turn sound on or off. Kling 3.0 Turbo returns video without sound in the current route. Topaz Astra does not generate sound — it enhances a finished clip.
Which new model is the most affordable?
Seedance 2.0 Mini — the most affordable tier of Seedance 2.0, built for volume and speed.
What is Topaz Astra and how do I use it?
Astra 2 is a video enhancement and restyle tool. Upload a finished clip, optionally add a prompt, choose 1080p or 4K, and tune creativity, sharpness, and realism.
