Veo 3.1
by Google DeepMind · Jan 2026
The most complete video generation system. Native audio, ingredients-to-video, style matching, character consistency, scene extension, object manipulation, camera/motion controls. State-of-art on MovieGenBench. The Swiss army knife of AI video, but at a premium.
Specifications
Max Resolution
1080p and 4K
Max Duration
8s base (extendable via scene extension)
FPS
24-30
Native Audio
Yes
ComfyUI Support
No
Fine-tunable
No
Min VRAM
Cloud only
Cost / Second
$0.20
Architecture
Proprietary (state-of-art T2V, I2V, T2A+V)
Parameters
Undisclosed
Inputs
T2V, I2V (ingredients-to-video), Style Reference, Character Reference, Scene Extension, First+Last Frame, Outpainting, Add/Remove Object, Camera Controls, Motion Controls, Character Controls (body/face/voice drive)
License
Commercial (subscription)
Strengths & Trade-offs
Strengths
- Best native audio (dialogue + SFX + ambient + music, generated natively in same pass). State-of-art T2V per Meta MovieGenBench. Ingredients-to-video (1-3 reference images for scene/character/object). Style reference (match aesthetic from reference image). Character consistency across scenes. Scene extension with visual+audio consistency. First+last frame transitions. Outpainting for aspect ratio adaptation. Add/remove objects with physics-aware placement. Camera controls (dolly, zoom, pan). Motion controls (draw object paths). Character controls (body+face+voice drive animation). 1080p and 4K output. SynthID watermarking.
Weaknesses
- 8s base clips (needs scene extension for longer). Most expensive per second ($0.20). Short speech segments still being refined. Cloud-only (no self-hosting). No open weights. Limited to Google ecosystem (Gemini, Flow, AI Studio, Vertex AI).
Best For
- Audio-native cinematic content. Dialogue scenes with natural lip sync. Sound design-heavy pieces. Cinematic one-takes with ambient audio. Style-matched content (reference image). Character-consistent series. Professional 4K deliverables. VFX (add/remove objects, outpainting).
Scores
Workflows on Floyo
Compare with other models