Sora 2 Text-to-Video
Generate 4–20s Sora 2 text-to-video clips with synchronized audio at 720p. Fixed 200–1,000 credits per clip, no waitlist, right in your browser.
Sora 2 Text-to-Video turns a written prompt into a short, physically coherent clip with synchronized audio. It is OpenAI's second-generation video model, and its strengths are believable object motion, sharp realism and strong instruction-following rather than cheap iteration. On Conut.ai it generates 4, 8, 12, 16 or 20-second clips at 720p in portrait or landscape, and pricing is fixed per clip instead of per second: 200 credits for 4 seconds, up to 1,000 for 20.
- Duration: 4s – 20s
- Resolution: 720p
- Price: 200–1,000 credits/clip
- Audio: Synchronized
How to use Sora 2 Text-to-Video
Write the shot, choose the frame, then pay for the duration you actually need. The 4-second tier exists so you can test the idea before committing.
- Write the prompt as a shot description Name who or what is on screen, one action, one camera move, the light and the mood. Sora 2 follows cinematic vocabulary such as dolly in or shallow depth of field.
- Add sound cues tied to visible action Audio is generated with the picture. Mention rain on metal, footsteps on gravel or distant traffic so the soundtrack has something concrete to build from.
- Choose portrait or landscape There are only two sizes, both 720p. Decide the destination format now, because the framing language in your prompt should match it.
- Pick the shortest duration that tells the beat Four seconds for a test, eight or twelve for a social clip, sixteen or twenty for a build-up. The price steps up with every tier.
- Check physics and sync before accepting Watch hands, feet and moving objects for weight, and confirm the sound lands with the action. If the shot fails, revise the prompt rather than extending it.
Settings reference
Sora 2 Text-to-Video has a deliberately small control surface: a prompt, a duration tier and an orientation. Everything else is decided by the model.
| Setting | Value | Notes |
|---|---|---|
| Prompt | Required | Describe the subject, action, camera behavior, lighting and the sounds tied to the action. |
| Duration | 4s / 8s / 12s / 16s / 20s | Five fixed tiers; there is no continuous duration slider. |
| Orientation | Portrait 720×1280 or landscape 1280×720 | Both output 720p. There is no square, 4:3 or ultra-wide option. |
| Price | 200–1,000 credits per clip | 200, 400, 600, 800 and 1,000 for the five tiers. Billed per clip, not per second. |
| Audio | Synchronized with the picture | Ambience and effects match the visible action. This is not a scripted-dialogue tool. |
| Reference inputs | None | The text endpoint accepts a prompt only; use Sora 2 Image-to-Video to animate a still. |
| Seed & negative prompt | Not exposed | Reproducibility comes from reusing the same prompt, not from a fixed seed. |
What Sora 2 text-to-video does well
Sora 2 is built around physical plausibility. Objects keep their weight, liquids behave like liquids, and camera moves feel like a real camera rather than a warp of the frame. Compared with earlier text-to-video models, it holds a scene together for longer and needs less prompt babysitting to avoid melting faces or sliding feet.
The model also generates sound with the picture instead of leaving the clip silent. Footsteps, wind, traffic, room tone and impacts arrive in sync with the action, which makes the output feel finished without a separate sound pass. That combination, coherent motion plus synchronized audio, is the reason to pay Sora 2 prices rather than draft on a cheaper model.
- Accurate physics and weight in moving objects
- Synchronized ambience and effects
- Strong prompt adherence and steering
- A wide stylistic range, from photoreal to animated
- 720p portrait or landscape output
Sora 2 duration and credit pricing
Sora 2 is billed per clip, not per second or per resolution. That makes the cost predictable but also means long clips carry a fixed premium. The 4-second tier is the cheapest way to test whether a prompt, camera move or style works before committing to a longer take.
- Fixed per-clip pricing, not per-second billing
- 720p output only on this endpoint
- Portrait (720×1280) or landscape (1280×720)
- No negative prompt or seed controls exposed
| Duration | Credits | Best for |
|---|---|---|
| 4s | 200 | Prompt and motion tests |
| 8s | 400 | Social clips and product beats |
| 12s | 600 | Short narrative scenes |
| 16s | 800 | Longer cinematic builds |
| 20s | 1,000 | Trailers and hero shots |
Sora 2 vs Seedance 2.5 vs WAN 2.6 vs Kling O3 Pro
Sora 2 competes on motion quality and audio, not on resolution or price per second. Seedance 2.5 and Wan 3.0 reach 4K and 1080p respectively; WAN 2.6 is far cheaper per second at 1080p; Kling O3 Pro offers granular 3–15 second durations. Choose Sora 2 when the shot needs its specific look and synchronized sound, and choose the others when resolution, duration or budget is the constraint.
- Choose Sora 2 for its motion style and synchronized audio.
- Choose Seedance 2.5 when 4K or a 30-second take is required.
- Choose WAN 2.6 when 1080p cost per second matters most.
- Choose Kling O3 Pro when the clip must land on an exact duration.
| Model | Resolution | Duration | Price | Audio | Best for |
|---|---|---|---|---|---|
| Sora 2 T2V | 720p | 4–20s | 200–1,000 /clip | Synchronized | OpenAI motion and sound |
| Seedance 2.5 T2V | 480p–4K | 5–30s | 60–600 /s | Native | 4K and long takes |
| WAN 2.6 T2V | 720p / 1080p | 5–15s | 30–40 /s | Native + upload | Cheap 1080p |
| Kling O3 Pro T2V | Endpoint-set | 3–15s | 80–100 /s | Optional | Fine duration control |
| Wan 3.0 T2V | 480p–1080p | 2–30s | 23–90 /s | Optional | Flexible length with audio |
Writing prompts that Sora 2 follows
Write the prompt like a short shot description: who or what is on screen, what happens, how the camera behaves, and what the light looks like. Sora 2 handles cinematic vocabulary well, so terms such as dolly in, tracking shot, shallow depth of field and volumetric light are useful rather than decorative.
Because the model generates audio, add one or two sound cues tied to visible action. Keep them concrete, such as rain on metal or footsteps on gravel. This endpoint is not a dialogue tool; asking for a scripted conversation is more likely to produce vague ambience than clean speech.
- One subject, one action, one camera move per clip
- Describe light and mood instead of stacking adjectives
- Add one or two sound cues tied to visible action
- Draft at 4 seconds, then move the winning prompt to 8 or 12
- Avoid cuts inside a clip; generate separate shots instead
Limitations and when to use another model
The clearest limitation is 720p output. If the final deliverable needs 1080p or 4K, Sora 2 is the wrong endpoint regardless of how good the motion looks. Fixed per-clip pricing also becomes expensive at 16 and 20 seconds compared with per-second models at the same length.
There is no negative prompt or seed, so reproducibility depends on reusing the same prompt and accepting variation. For scripted dialogue or a specific voice track, use WAN 2.5 image-to-video. For cheap silent drafts, WAN 2.2 Spicy costs far less per second.
- 720p maximum resolution
- Fixed per-clip pricing; compare at your target duration
- No seed or negative prompt
- Audio is synchronized ambience, not scripted dialogue
- For lip sync with a chosen voice, use WAN 2.5
Prompt examples for Sora 2 Text-to-Video
- Rainy street at night: A lone figure walks toward the camera through a rain-soaked city street at night, neon signs reflecting on wet pavement, slow dolly in, shallow depth of field, cinematic color grade. Rain on metal, distant traffic, low synth hum. One action, one camera move, two sound cues.
- Product reveal: A matte black wireless speaker sits on a dark reflective surface as the camera orbits slowly; a single spotlight sweeps across the mesh grille, fine dust drifting in the beam. Soft mechanical click, deep room tone, luxury commercial lighting. Keep the product still and let camera and light create motion.
- Mountain sunrise: Morning fog rolls over a mountain ridge as the first sunlight breaks through pine trees; camera pushes forward slowly, volumetric light rays, cold green and amber palette. Wind in the pines, distant birds, quiet morning ambience. Landscapes tolerate large camera moves without a subject to warp.
- Sci-fi corridor: A figure stands in a dark sci-fi corridor as overhead lights activate one by one toward a sealed door; camera tracks forward past the subject, steam venting from pipes, cool blue and amber lighting. Electrical hum, distant alarm chirps. Use environmental motion to add energy without moving the subject much.
- Food commercial: A chef drops a seared steak onto a hot cast-iron pan, oil sizzling and smoke curling upward; macro camera slowly pushes in, warm tungsten light, shallow depth of field. Sharp sizzle, kitchen room tone, subtle flame whoosh. Physical interactions are where Sora 2's physics simulation shows.
- Trailer moment: A character stands on a rooftop as a storm gathers, coat whipping in the wind, lightning illuminating the skyline behind them; camera tilts up from boots to face, dramatic clouds, teal and orange grade. Thunder, wind, rain on concrete. Use a 16 or 20-second tier to give the build-up room.
Pro tips for Sora 2 Text-to-Video
Sora 2 is priced per clip. The skill is deciding when a 200-credit test is enough and when a longer tier is justified.
- Test every idea at 4 seconds The 200-credit tier is the cheapest way to check composition, motion and audio. Move the winning prompt to 8 or 12 seconds after it reads.
- Do the per-second math at 20 seconds A 20-second clip costs 1,000 credits, which works out to 50 per second. Compare that against per-second models when your shot really needs the length.
- One camera move per clip Sora 2 handles a single clear move well. Stacking a dolly, a pan and a tilt in one prompt makes the physics fight itself.
- Do not ask for scripted dialogue The audio track is synchronized ambience and effects. For a specific voice or lip sync, use WAN 2.5 Image-to-Video instead.
- Use 720p as a creative constraint There is no 1080p or 4K here. Compose for the web and social, and move to WAN 2.6 or Seedance 2.5 when the deliverable needs more pixels.
Frequently asked questions
How much does Sora 2 text-to-video cost?
Pricing is fixed per clip: 200 credits for 4 seconds, 400 for 8, 600 for 12, 800 for 16 and 1,000 for 20 seconds.
How long can a Sora 2 clip be?
Up to 20 seconds. The endpoint supports 4, 8, 12, 16 and 20-second durations.
What resolution does Sora 2 output?
720p in either portrait (720×1280) or landscape (1280×720). There is no 1080p or 4K option on this endpoint.
Does Sora 2 generate audio?
Yes. It generates synchronized ambience and effects that match the visible action. It is not a scripted dialogue or lip-sync tool.
Is a prompt required?
Yes. Describe the subject, action, camera movement, lighting and any sound cues you want in the clip.
Can I set a seed or negative prompt?
No. This endpoint does not expose seed or negative-prompt controls. Reuse the same prompt for a similar result, but expect variation.
What is the difference between Sora 2 text-to-video and image-to-video?
Text-to-video builds the scene from the prompt alone. Image-to-video starts from your reference image and preserves its subject, lighting and composition while adding motion and sound.
Does Sora 2 support 9:16 vertical video?
Yes. Choose the portrait size (720×1280) for vertical output, or the landscape size (1280×720) for 16:9.
Compare with other text-to-video models
- Sora 2 Image-to-Video
- Seedance 2.5 Text-to-Video
- WAN 2.6 Text-to-Video
- Kling O3 Pro Text-to-Video
- Wan 3.0 Text-to-Video
- Browse all AI video models
Specifications
- Provider: Wavespeed
- Model type: Text to video
- Credit pricing: 200 credits per video
- Capabilities: new, cinematic, realistic, audio-sync, multi-shot
About Conut.ai
Conut.ai is an AI image, video and audio generation platform with 50+ creative tools. It supports text-to-image, image-to-video, face swap, motion control, relighting, inpainting and video effects through models including Nano Banana Pro, Seedream, GPT Image, Sora 2, Kling, WAN and Seedance. Start free — no credit card required.