Model directory
The AI models that matter for film.
The video, image and audio models Cinemagiq runs, with real examples, what each is good at and when to reach for it. Well-known models we don't run are listed too, with the closest alternatives.
Video models
Seedance 2.5
ByteDanceUp to 30 seconds of video from a prompt, frames or references
View detailsSeedance 2.0
ByteDanceImage-to-video with end frames, references and optional sound
View detailsKling O3
KuaishouImage-to-video, video edits and motion control from Kuaishou
View detailsKling 4.0
Kuaishou30-second clips with sound, from text, frames or up to 10 references
View detailsMiniMax H3
MiniMaxText or frames to video with sound, plus sharp 2K and 4K upscales
View detailsMiniMax H3 Max
MiniMaxFaster MiniMax video with tighter prompt adherence and sound
View detailsHailuo 2.3
MiniMaxCinematic motion from a prompt or a first frame, in 6 or 10 seconds
View detailsGemini Omni 1.1 Flash
Google DeepMindVideo with dialogue and ambience, from 360p drafts to 4K
View detailsGrok Imagine Video 1.5
xAI1 to 15 second clips with native audio from a prompt or references
View detailsHappy Horse 1.1
AlibabaAnimate a still up to 1080p, plus prompt-based video edits
View detailsPixVerse C1
PixVerseFast, low-cost image-to-video from 1 to 15 seconds
View detailsRunway Aleph 2.0
RunwayEdit footage by prompt or keyframes, and transfer real performances
View detailsOther video models (not in Cinemagiq)
Veo 3.1
Google DeepMindGoogle DeepMind's video model with sound built in
Not available in CinemagiqView detailsRunway Gen-4.5
RunwayRunway's video generation model for creative teams
Not available in CinemagiqView detailsPika
PikaA fast, playful AI video platform for quick ideas
Not available in CinemagiqView detailsLuma Dream Machine
Luma AILuma AI's video app, built on its Ray video models
Not available in CinemagiqView detailsWan
AlibabaAlibaba's AI video model family, with open-weight releases
Not available in CinemagiqView detailsLTX-2
LightricksLightricks' open-weight video model with synchronized sound
Not available in CinemagiqView detailsImage models
Nano Banana 2.1
Google DeepMindGoogle's newest Nano Banana and the default image model in Cinemagiq
View detailsNano Banana 2
Google DeepMindGoogle's Nano Banana 2 for stills up to 4K, plus Nano Banana 2 Lite
View detailsNano Banana Pro
Google DeepMindGoogle's Pro tier Nano Banana: stills up to 4K with six references
View detailsGPT Image 2.5
OpenAIOpenAI's GPT Image 2.5: Flare for speed, Sunburst for precision
View detailsGPT Image 2
OpenAIOpenAI's GPT Image 2 and GPT Image 1.5 in 1:1, 3:2 and 2:3 frames
View detailsSeedream 5.0
ByteDanceByteDance's Seedream 5.0 in Pro, Lite and Flash, at exact ratios
View detailsFLUX.2
Black Forest LabsFLUX.2 Pro, Max and Dev: stills that respond to lens and camera terms
View detailsQwen Image 3
Alibaba (Qwen)Alibaba's Qwen Image 3, for stills where the words have to read right
View detailsRecraft V4.1
RecraftRecraft V4.1 and Recraft V4.1 Pro: prompt-only stills at 1K or 2K
View detailsGrok Imagine 2.0
xAIxAI's Grok Imagine 2.0 image model, at 1K or 2K and two quality levels
View detailsOther image models (not in Cinemagiq)
Audio models
Upscalers
Head-to-head comparisons
Not sure which model to pick? These matchups break it down shot by shot.
- Seedance 2.5 vs Kling 4.0
- Nano Banana Pro vs GPT Image 2.5
- Kling O3 vs Veo 3.1
- Kling O3 vs Hailuo 2.3
- Kling O3 vs Runway Gen-4.5
- Veo 3.1 vs Runway Gen-4.5
- Veo 3.1 vs Hailuo 2.3
- Seedance 2.5 vs Hailuo 2.3
- Wan vs Kling O3
- Pika vs Luma Dream Machine
- Pika vs Runway Gen-4.5
- GPT Image 2 vs Nano Banana 2
- FLUX.2 vs Midjourney
- FLUX.2 vs GPT Image 2
Learning a new model? Read the prompt guides for structures that work and real examples.