Google
7.0/10
- Realistic motion and prompt adherence with native synchronized audio
- Strong for cinematic scenes and ambience out of the box
- Availability and pricing vary by platform and SKU
- Generation still limited to relatively short clip lengths
MiniMax
MiniMax Hailuo 2.3
4.4/10
- Maker
- MiniMax
- Origin
- China
- Released
- Oct 2025
- Category
- Video Generation
Strengths & weaknesses
- Strong motion quality for short cinematic clips
- Competitive pricing for social and short-form content
- Best suited to short clips rather than longer narrative sequences
- Niche provider with a smaller support ecosystem
Evaluation
100% of weight measuredScored on Hailuo 2.3 · sources as of Nov 4, 2026
Confidence band 4.4–4.7 · a model inside this range isn’t meaningfully apart from this one
- Text-to-videow603.6
Rule B1,206 rating (95% 1,200–1,212) from 89,138 votes · Scale 1,000–1,550, 0 to 10 · LMArena text-to-video arena (hailuo-2.3) · Sep 22, 2026
- Image-to-videow405.6
Rule B1,262 rating (95% 1,257–1,267) from 405,786 votes · Scale 950–1,500, 0 to 10 · LMArena image-to-video arena (hailuo-2.3) · Sep 22, 2026
w = weight, each criterion’s share of the overall score. Missing marks don’t count for or against. How scoring works
More in Video Generation
View all →←OpenAI
Under evaluation
- Set the early bar for realistic, physically-consistent video generation
- Strong brand recognition and broad public familiarity
- OpenAI closed the Sora app in April 2026 and removes the API on September 24, 2026
- Superseded by newer models from Google, ByteDance, and Kuaishou
Kuaishou
Under evaluation
- Excellent character motion and dramatic, controllable camera moves
- Strong image-to-video animation from a single source frame
- Frame preservation needs careful checking on longer generations
- Native audio exists but is often disabled in third-party integrations
Runway
Under evaluation
- Full creative production environment, not just a generation endpoint
- Strong iteration, editing, and referencing tools for professional workflows
- Audio still has to be added separately in most pipelines
- Steeper learning curve than simple prompt-and-generate tools