Why choose it
- Always-on native audio
- Animate a starting image
- 4–20 second durations
Original model creator
OpenAI
Sora 2
OpenAI Sora 2 for 720p clips with always-on native audio from text or a starting image.
Loading Sora 2 sample
Official sampleSourced output
Official sample · Results vary by prompt and settings
See the model move.
Direct answer
When should you use Sora 2?
Sora 2 is OpenAI’s video model in Video Lab. Generate from a prompt or animate a starting image at 720p with native audio. Durations are 4, 8, 12, 16, or 20 seconds. Use Sora 2 Pro when you need true 1080p.
Best for
OpenAI motion, native-audio social clips, and 720p shots from text or a still.
Technical readout
The supported shot envelope.
- Duration
- 4, 8, 12, 16, or 20 seconds
- Resolution
- 720p
- Aspect ratio
- 16:9 or 9:16
- Audio
- Native audio
What to consider
- 720p only
- No first/last-frame or multi-image reference modes
Prompt direction
How to prompt Sora 2
Write spoken lines and camera language together. Sora generates picture and sound in one pass, so dialogue and ambient noise belong in the same prompt.
Start a Sora 2 promptQuestions, answered
Sora 2 FAQ
What is Sora 2 best for?
Sora 2 is OpenAI’s 720p video SKU for native-audio clips from text or a starting image.
Can Sora 2 use first and last frames?
No. Video Lab can animate one starting image. First-to-last-frame and reference-image modes are not available.
Does Sora 2 generate audio?
Yes. Native audio is always generated with the clip.
Change the lens
Compare Sora 2 with another model.
Official sampleOpenAIOriginal model creatorSora 2 Pro
OpenAI Sora 2 Pro for 720p or true 1080p clips with always-on native audio from text or a starting image.See model
Official sample
GoogleOriginal model creatorVeo 3.1
A premium text-and-image video model for cinematic shots, guided frames, references, optional audio, and up to 4K output.See model
Official sampleGoogleOriginal model creatorVeo 3.1 Fast
A faster, lower-cost Veo 3.1 SKU for cinematic shots, guided frames, references, optional audio, and up to 4K.See model
Official sample
ByteDanceOriginal model creatorSeedance 2.5
A next-generation text-and-image video model with native audio, 30-second shots, first/last frames, and deep reference control.See modelOpenAI lens selected
Direct your first Sora 2 shot.
Write the prompt, choose supported controls, and generate.
Open Sora 2 in Video Lab