Sora 2 prompting cheatsheet (motion, camera, audio cues that actually work)
Production-tested Sora 2 prompt structure: subject + action + camera + lens + lighting + duration + audio. Plus the cues Sora 2 respects vs ignores.
Seven-slot prompt template
Sora 2 follows natural language closely. Order matters less than completeness — every slot should be present.
| Item | Description | Example |
|---|---|---|
1. Subject | Who or what is in frame. | A 60-year-old mechanic with grease-stained hands |
2. Action | What they're doing. Use a present participle. | carefully tightening a bolt on an engine block |
3. Camera move | Dolly, push-in, orbit, handheld, locked-off, crane. | slow orbit camera move counterclockwise around the subject |
4. Lens / framing | Wide, medium, close-up, macro. | medium shot, 35mm equivalent, eye level |
5. Lighting | Source, direction, quality. | warm tungsten overhead light, soft shadows |
6. Duration | Sora 2 supports up to 60s; specify if you want a long take. | 12-second continuous take, no cuts |
7. Sound design | Ambient + foley. Sora 2 does sound but not dialogue. | subtle metallic clank, distant garage hum |
Cues Sora 2 respects
| Item | Description | Example |
|---|---|---|
Camera direction | Push-in, orbit, dolly out — usually executed precisely. | |
Lighting source | Window left, key from below, hard sun — strongly respected. | |
Lens length | Wide vs telephoto framing measurably changes output. | |
Physical materials | Wet, dusty, oily, foam, dust — Sora 2 simulates physics well. | |
Cues Sora 2 ignores (do not rely on these)
| Item | Description | Example |
|---|---|---|
Synchronised dialogue | Sora 2 will not lip-sync spoken lines in 2026. Use Veo 3 if you need speech. | |
Exact text in scene | Signs and screens often produce nonsense; pick frames where text is incidental. | |
Brand logos | Specific brand logos rarely render correctly; describe generically. | |
Counts above 7 | 'A crowd of 20 people' becomes 'a crowd'. Don't expect exact counts above a handful. | |
Settings
| Item | Description | Example |
|---|---|---|
Resolution | 1080p default; 4K via post-upscale. Higher resolution adds cost, not adherence. | |
Aspect ratio | 16:9, 9:16, 1:1 supported natively. Decide before generating — recropping degrades quality. | |
Seed | Pin the seed when iterating phrasing. | |
FAQ
Can Sora 2 generate dialogue?
Not in 2026. Sora 2 generates ambient sound and foley but not synchronised lip-synced dialogue. Use Veo 3 if you need talking characters.
What's the max clip length on Sora 2?
60 seconds for a single continuous clip at 1080p. Longer compositions are achieved by chaining multiple clips.
Does Sora 2 follow camera direction prompts?
Yes — push-in, orbit, dolly, handheld, locked-off — Sora 2 respects camera cues better than any other video model in 2026.
Last updated: 2026-06-01.