Générez des vidéos cinématiques de 15s avec séquences multi-plans, contrôle du premier/dernier cadre, 21:9 et audio natif — le tout en 1080P.
Real outputs from our production — not cherry-picked. Hover any clip to play, click the speaker icon to hear the audio.
Hover to play · Click speaker to unmute
Real outputs from our production — not cherry-picked. Hover any clip to play, click the speaker icon to hear the audio.
Monde alien fantastique
Publicité parfum de luxe
Teaser de lancement
Publicité voyage Maldives
Duel wuxia oriental
Scène de film de courses
Vaisseau vers les étoiles
Golden Retriever et papillons
Chef de sushi à l'œuvre
Robe rouge sur la falaise
Duel de minuit
Vlog POV petite amie
Gros plan de cuisine alléchante
Danse urbaine de rue
Publicité rouge à lèvres de luxe
The first AI video model with true multi-shot cinematography. Think of it as an AI director, not a single-shot generator.
Write "Camera switch:" in your prompt and Seedance 2 renders a new shot — wide, medium, extreme close-up. A single prompt yields a 4-shot cinematic sequence.
Upload two keyframes, Seedance 2 fills the motion between. Perfect for ad transitions, product reveals and storytelling arcs.
True cinema aspect ratio. Banners, trailers, anamorphic shots — no other major model supports 21:9.
Optional. Seedance 2 can add ambient tones, subtle music or synchronized sound effects. One generation, fully scored.
Feed up to 9 reference images, 3 reference videos and 3 audio tracks into a single text-to-video prompt. Style, character and pace — all controllable.
Any length from 5 to 15 seconds in 1-second steps. 480P / 720P / 1080P — pick the quality that fits your workflow.
Seedance 2 works like a film set — you write the shot list, it handles the cinematography.
Describe each shot with a time code. Use "Camera switch:" between shots for Seedance 2's multi-shot syntax.
Seedance 2 plans cinematography, lighting, motion, then renders every shot with style consistency. Takes ~2 minutes.
Download the finished cut with optional native audio. Commercial-ready, no watermark, no cleanup needed.
How does Seedance 2 stack up against the competition? Side-by-side on the capabilities that matter for production.
| Capability | Seedance 2 | Veo 3 | Runway Gen-4 | Kling 2.5 |
|---|---|---|---|---|
| Max duration | 15s | 8s | 10s | 10s |
| Max resolution | 1080P | 1080P | 1080P | 1080P |
| First & last frame | Yes | Yes | No | Yes |
| Multi-shot switching | Yes | Partial | No | Partial |
| Native audio generation | Yes | Yes | No | No |
| 21:9 widescreen | Yes | No | No | No |
| Multimodal reference | Image + Video + Audio | Image | Image | Image |
Data as of April 2026. Seedance 2 is the only major video model supporting 21:9 cinematic widescreen and multimodal (image + video + audio) reference inputs.
Seedance 2 excels at cinematic storytelling — the scenarios that need true directorial control, not just single-shot generation.
10-second teasers with dramatic reveal, close-up, lifestyle and logo lock. One prompt, four shots, ready to run on TikTok and Instagram.
TikTok, Reels, Shorts. 9:16 vertical, 8-15 seconds, snappy cuts, captions baked into the prompt — ready to upload.
Epic establishing shots, action chases, hero close-ups — multi-shot sequences that tell a complete story in 10-15 seconds.
Aerial sweeps, underwater dives, lifestyle shots — destination content that ranks on social feeds and travel sites.
Seedance 2 responds to directorial language. A few specific syntax tricks unlock the full multi-shot power.
Think like a director, write like a script
Seedance 2 recognizes "Camera switch:" as a hard cut to a new shot. Stack multiple switches for multi-shot sequences.
Example
"Shot 1: Wide aerial of stormy sea.\nCamera switch:\nShot 2: Close-up on warship's bow cutting through waves.\nCamera switch:\nShot 3: Hero's face at the helm."
Bracket notation like [0:00-0:02.5] tells Seedance 2 exactly how long each shot should last.
Example
"[0:00-0:03] Shot 1 — Establishing\n[0:03-0:05.5] Shot 2 — Close-up\n[0:05.5-0:08] Shot 3 — Hero beat"
Each shot should include: Scene, Subject, Action, Camera, Lighting, Mood. Missing one often causes the AI to improvise.
Example
"Scene: neon alleyway at night. Subject: hooded figure. Action: running toward camera. Camera: low-angle tracking. Lighting: magenta rim. Mood: tense."
Seedance 2 doesn't understand "stunning" or "epic". Replace them with concrete visual details.
Example
"Avoid: "stunning epic landscape". Prefer: "golden-hour lagoon with jade reefs and a silver wake trailing behind a small boat"."
Seedance 2 supports 5-15 seconds in 1-second steps, aspect ratios from 21:9 to 9:16. State both early in your prompt.
Example
"A cinematic 10-second, 21:9 widescreen chase sequence through a neon canyon"
Upload a start image and an end image, Seedance 2 fills the motion between. Ideal for product reveals or dramatic transformations.
Example
"First frame: closed leather wallet. Last frame: wallet open showing cards. Prompt: "smooth flip-open motion, soft studio light"."