Ten models, every spec that matters. Resolution, duration, audio, reference-to-video, pricing — and which one to use when.
| Seedance 2.5ByteDance | Seedance 2.0ByteDance | Grok Imagine 1.5xAI | Flux 3Black Forest Labs | Gemini OmniGoogle DeepMind | Kling V3 4KKuaishou | Wan 3.0 Prime ProAlibaba | MiniMax H3MiniMax | MiniMax H3 MaxMiniMax | LTX 2.5 ProLightricks | |
|---|---|---|---|---|---|---|---|---|---|---|
| Video Specs | ||||||||||
| Max Resolution | 1080p | 4K (2160p) ★ | 1080p | 1080p | 4K (2160p) ★ | 4K (2160p) ★ | 4K (2160p) ★ | 2K | 1080p | 1080p |
| Max Duration | 30s ★ | 15s | 15s | 20s | 10s | 15s | 30s ★ | 15s | 15s | 10s |
| Frame Rate | 24 fps | 24 fps | 24 fps | 24 fps | 24 fps | 24 fps | 24 fps | 24 fps | 24 fps | 24–50 fps (fal) ★ |
| Aspect Ratios | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16 | auto, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, 9:21 ★ | 16:9, 9:16 | 16:9, 9:16, 1:1 | 16:9, 9:16, 1:1, 4:3, 3:4 | 16:9, 21:9, 4:3, 1:1, 3:4, 9:16 | 16:9, 21:9, 4:3, 1:1, 3:4, 9:16 | 16:9, 9:16 |
| Input Modes | ||||||||||
| Text to Video | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Image to Video | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ (via R2V) | ✓ | ✓ | ✓ | ✓ |
| Start + End Frame | ✓ | ✓ | ✗ | ✓ (dedicated FLF) | ✓ (fal) | ✗ | ✓ | ✗ | ✓ (fal) | ✓ (fal) |
| Reference to Video | Flat @Image refs (up to 30) ★ | Flat @Image refs | Flat @Image refs | ✗ | @ refs (up to 3) | Structured elements | Flat @ refs (10 img / 5 video / 5 audio) | @ refs (10 img / 5 audio / video) | @ refs (10 img / 5 audio / video) | ✗ |
| Audio Reference | ✓ (up to 10) ★ | ✓ (3) | ✗ | ✗ | ✗ | ✗ | ✓ (5) | ✓ (5, paired) | ✓ (5, paired) | ✗ |
| Audio & Camera | ||||||||||
| Native Audio | ✓ | ✓ | ✓ Always on | ✓ Always on | ✓ Always on | ✓ | ✓ | ✓ Always on | ✓ Always on | ✓ |
| Camera Control | Good | Good | Good | Good | Good | Precise ★ | Good | Good | Good | Good |
| FPS Selector | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ | 25 / 50 ★ |
| Character Consistency | @ refs (up to 30) ★ | @ reference | @ reference | Limited | @ reference (3) | R2V elements | @ refs (img + video) | @ refs (img + video) | @ refs (img + video) | ✗ |
| Quality | ||||||||||
| Physics Realism | Industry-leading ★ | Excellent | Good | Good | Strong | Excellent | Excellent | Strong | Strong | Good |
| Color Science | Strong | Strong | Photorealistic | Photorealistic ★ | Good | Cinematic | Strong | Good | Good | Neutral |
| Lip Sync | Good ★ | Limited | Limited | Limited | Good | Limited | Limited | Limited | Limited | ✗ |
| Pricing (via Venice.ai BYOK, live quotes) | ||||||||||
| ~10s clip (720p) | $2.90 | $1.90 | $1.61 | $1.88 | $1.16 ★ | — | $5.06 (2k) | $1.00 (768p) | $0.24 (768p) | $1.20 |
| ~10s clip (1080p) | $5.10 | $4.70 | $2.88 | $3.20 | $1.75 | — | $4.74 | — | $0.48 ★ | $1.70 |
| ~10s clip (4K) | — | $9.70 | — | — | $3.50 | $4.62 | $5.40 | $1.63 (2K) | — | — |
| Audio Toggle | Optional | Optional | Always on | Always on | Always on | Optional | Optional | Always on | Always on | Optional |
| Open Source | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ | Open weights (geo-restricted) | Open weights (geo-restricted) | Open weights ★ |
| Arena Rank | — (new) | Top 5 | Top 5 | — | #1 T2V + I2V | Top 5 | — (new) | #2 overall, #1 editing ★ | — (new) | — |
Thirty-second generations, the longest in the field, with up to 30 image references and 10 audio references per clip. It trades 2.0's 4K for double the duration. The serialization engine.
4K output, industry-leading physics, start and end frame control, and audio references. The most controllable generation in the table, and still the motion benchmark.
Native 4K with the most precise camera control available. Premium price, premium frame. If the output needs to hold up on a big screen, start here.
The only pairing of 4K output and 30-second generations in the field, with start and end frame control and optional native audio. Premium tier, premium price.
The editing family's budget tier: $0.48 for ten seconds at 1080p — by far the cheapest 1080p in the table — with image, video, and audio references.
Fastest generation times in the field. Seven aspect ratios, 1080p, 15 seconds, audio always on. For social volume and rapid prototyping, nothing ships faster.
A dedicated first-last-frame variant, eight aspect ratios including auto, and the most photographic color science in the table. The stills-photographer's video model.
1080p at up to 50fps with synchronized audio for $1.70–1.90 per ten seconds. Open weights, 10-second generations, FPS selector on fal. Self-hostable for zero recurring cost.
CinePrompt supports BYOK (bring your own key) generation across multiple providers. Pick your provider, paste your key, and generate directly inside the prompt builder.
Privacy-first platform carrying every model on this page: Seedance 2.5 and 2.0, Grok Imagine 1.5, Flux 3, Gemini Omni, Kling V3 4K, Wan 3.0 Prime Pro, MiniMax H3 + H3 Max, LTX 2.5, and more. No data logging, no content filters. Pro subscription or API access.
The developer's pick. Access Seedance 2.5 and 2.0, Grok Imagine 1.5, Gemini Omni, Kling V3, LTX 2.5, and hundreds more via a single API. Pay-per-second pricing, no subscription required.
Seedance, Kling, WAN, Nano Banana Pro, and 40+ more at competitive pricing. The budget-conscious pick for high-volume generation without sacrificing model quality.
CinePrompt generates model-specific prompts for every major model family on this page — plus 50 more. Pick your model and get output tuned to what it actually understands.
Open Prompt Builder →