Up to 30 seconds in one take
According to ByteDance, Seedance 2.5 creates up to 30 seconds in a single generation, against 15 for the Seedance 2.0 series. Enough for a full ad or scene without stitching several clips together.
Video model · ByteDance Seed
Seedance 2.5 — Videos up to 30 seconds with sound, guided by your references.
Seedance 2.5 is the video model from ByteDance’s Seed team, introduced in July 2026 as the successor to the Seedance 2.0 series. On Deepnia it creates 4–30 second videos in a single generation, with voices, sound effects and music generated alongside the picture, from text or from up to 16 reference images and videos, in 480p or 720p.
- 4–30 s
- Sound always on
- 16 references incl. 10 videos
- 480p or 720p
- 6 aspect ratios incl. 21:9
- 30,000-character prompt
The video tool opens with Seedance 2.5 already selected. The exact cost shows before every creation; you sign up when you generate.
- A prompt of up to 30,000 characters
- Up to 16 images and videos (optional)
- A 4–30 s video with sound
At a glance
- Maker
- ByteDance Seed
- Length
- 4 to 30 s (5 s by default)
- Resolution
- 480p or 720p (720p by default)
- Aspect ratios
- 16:9, 9:16, 1:1, 3:4, 4:3, 21:9
- References
- Up to 16 media, including 10 videos
- Sound
- Always on
- Prompt
- Up to 30,000 characters
What Seedance 2.5 does best
Sound in the same generation
Voices, sound effects and background music are generated with the picture: sound is always included. Put dialogue lines in quotation marks and describe the sounds at the end of the prompt.
16 references, including 10 videos
In Reference mode, combine up to 16 images and videos to guide a character, a product, a place, a style or a camera move. ByteDance says the model reads the intention, framing and cinematic language of reference videos.
Time markers to the second
Split the scene with markers such as “0-6 s” or “6-14 s”: ByteDance says Seedance 2.5 understands them. Prompts can run to 30,000 characters on Deepnia.
Directed like a film shoot
ByteDance highlights professional-style camera movement, precise blocking of performers in the frame and smoother motion.
How to use Seedance 2.5 on Deepnia
Open the video tool
The “Try Seedance 2.5” button opens Deepnia’s video tool with the model already selected.
Add your references
Stay with text only, or switch to Reference mode to add images and videos: 16 media in total, including 10 videos.
Write the scene
Split the action with time markers, describe the camera, then the sounds you want at the end of the prompt.
Set it up and generate
Set the length (4–30 s), resolution and aspect ratio, check the cost shown, then create.
Seedance 2.5 on Deepnia: what you can set
- Inputs
- Text only, Reference mode (images and videos), Keyframe mode (start frame, optional end frame)
- Length
- 4 to 30 s, to the second (5 s by default)
- Resolution
- 480p or 720p (720p by default)
- Aspect ratios
- 16:9, 9:16, 1:1, 3:4, 4:3, 21:9 (16:9 by default)
- References
- Up to 16 media including 10 videos; images up to 10 MB, videos up to 100 MB, 30 s of video in total
- Sound
- Always on: voices, sound effects and music
- Prompt
- Required, up to 30,000 characters
Prompt ideas to try
Copy a prompt, open the tool and adapt it to your project.
30-second ad for a restaurant
- 16:9
- 30 s
- 720p
- Reference: 3 photos (storefront and 2 dishes)
A 30-second ad in warm late-afternoon light. 0-6 s: the restaurant storefront from the first reference image, on a lively street, passers-by in the foreground. 6-14 s: slow close-ups of the dishes from the other two images, rising steam, fresh herbs. 14-22 s: a cook turns skewers on the grill, sparks and smoke. 22-30 s: a family shares the meal, laughing, as the camera gently pulls back. Sound: sizzling grill, street ambience, light instrumental music. No text or subtitles on screen.
Vertical lookbook for a fashion label
- 9:16
- 20 s
- 720p
- Reference: 3 outfit photos and 1 walking video
A 20-second vertical video for a ready-to-wear label. A model wears the three outfits from the reference images one after another, matching the walk and rhythm of the reference video. Setting: a rooftop at sunset with the city behind. Cuts on the beat, waist shots then full-length shots, fabric moving in the wind. Sound: instrumental electronic music and footsteps. No on-screen text.
Open-air concert teaser in Johannesburg
- 21:9
- 25 s
- 720p
- Text only
A 25-second cinematic teaser for an open-air concert in Johannesburg. 0-5 s: aerial view of the city at dusk. 5-12 s: crew members set up the stage, spotlights switch on one by one. 12-20 s: the crowd dances, arms raised, golden confetti. 20-25 s: slow-motion close-up of a drummer. Sound: building percussion, roaring crowd, one second of silence before the final drum hit. No subtitles.
Comedy sketch without dialogue for social media
- 9:16
- 15 s
- 480p
- Reference: 1 photo of the character
A 15-second vertical comedy sketch. The character from the reference image, who sells phone accessories in a covered market, proudly stacks boxes. 0-5 s: he places the last box. 5-10 s: one box slips. 10-15 s: the whole stack collapses and he stares into the camera in despair. Sound: market chatter, comic sound effects, short upbeat music. No dialogue, no subtitles.
Hotel video with an English voice-over
- 16:9
- 30 s
- 720p
- Text only
A 30-second video for a seaside hotel, aimed at international guests. Slow shots: palm trees at sunrise, a pool facing the ocean, a bright bedroom, a candlelit dinner on the beach. Calm male voice-over: "Wake up to the sound of the ocean." Sound: waves, birds, soft acoustic guitar. No on-screen text.
Training clip for a sewing workshop
- 16:9
- 20 s
- 720p
- Text only
A 20-second documentary-style training video. A tailor shows step by step how to attach bias binding to a neckline: wide shot of the workshop, close-ups of his hands and the sewing machine, then the finished garment worn by a customer. Natural window light. Sound: humming machine, snipping scissors, workshop ambience, no music. No on-screen text.
Tips for better results
- Split the video with time markers to the second (“0-6 s”, “6-14 s”): according to ByteDance, Seedance 2.5 follows them better than shot numbers.
- Describe the sounds at the end of the prompt; ByteDance also suggests a notation: (music), <sound effect>, {dialogue}.
- Name the language before each line of dialogue.
- For on-screen text or lines in French, keep sentences short and check the result; for longer dialogue in French, Gemini Omni handles the language best.
- Keep to 1–5 main subjects: that is what ByteDance recommends for a stable result.
- Validate the idea at 480p over a few seconds, then generate the full version at 720p.
- If you want no subtitles or music, say so in the prompt (“no subtitles”, “no music”).
Good to know
- To tell a story longer than 30 seconds, create several videos and join them in your editor.
- To keep a character or product identical from one video to the next, reuse the same reference images and describe it with the same words.
- Pick 9:16 for Reels, TikTok and Shorts, 16:9 for YouTube and 21:9 for a cinematic look.
- Your reference videos can add up to 30 seconds, 100 MB each: short, focused clips are enough to guide a movement or a framing.
Seedance 2.5 or another model?
Deepnia’s other video models, to pick the right one for your project.
| Model | Length | Resolution | Aspect ratios | Best for |
|---|---|---|---|---|
| Seedance 2.5This model | 4–30 s | 480p–720p | 16:9 · 9:16 · 1:1 · 3:4 · 4:3 · 21:9 | Videos up to 30 seconds with sound, guided by your references. |
| Seedance 2.0 | 4–15 s | 480p–1080p | 16:9 · 9:16 · 1:1 · 3:4 · 4:3 · 21:9 | 1080p (Standard variant), Mini for quick tests, up to 15 s. |
| Kling 3.0 | 3–15 s | Set by the mode | 9:16 · 1:1 · 16:9 | A 5-shot editor, persistent characters and optional sound. |
| Gemini Omni | 4, 6, 8 or 10 s | 720p–4K | 9:16 · 16:9 | 4–10 s clips up to 4K with 7 reference images. |
| HappyHorse | 3–15 s | 720p–1080p | 9:16 · 1:1 · 16:9 · 4:3 · 3:4 | Editing an existing 3–60 s video. |
| Grok Video | 6–30 s | 480p–720p | 9:16 · 1:1 · 16:9 · 2:3 · 3:2 | Also up to 30 s, from text or a single photo. |
Frequently asked questions
How long can a Seedance 2.5 video be?
30 seconds. On Deepnia you choose from 4 to 30 seconds, to the second (5 s by default); that is also ByteDance’s official range for one generation.
What resolution do the videos come in?
480p or 720p, 720p by default. Validate the idea at 480p over a few seconds, then generate the full version at 720p.
Does the video come with sound?
Yes, always: voices, sound effects and music are generated together with the picture. Describe the sounds you want at the end of your prompt.
Can I reuse the style or movement of a video?
Yes: in Reference mode, add up to 10 videos (30 seconds in total) and say in the prompt what to reuse: the camera move, the walk, the rhythm or the style. ByteDance says Seedance 2.5 reads the framing and cinematic language of reference videos.
How many references can I add?
Up to 16 media on Deepnia, including up to 10 videos and 30 seconds of video in total. ByteDance recommends keeping 1–5 main subjects for a stable result.
Seedance 2.5 or Seedance 2.0: which one should I use?
Seedance 2.5 to go beyond 15 seconds (up to 30), add more references (16 versus 7) and write very long prompts (30,000 characters versus 2,500). Seedance 2.0 for 1080p (Standard variant) or quick tests with the Mini variant. Both always generate sound.
How do I edit a video I already have?
With HappyHorse, another Deepnia video model: add your 3–60 second video and describe the change you want, such as a new style or a swapped product. Seedance 2.5 creates new videos of up to 30 seconds in one take.
When was Seedance 2.5 released?
ByteDance Seed introduced Seedance 2.5 on July 31, 2026. It succeeds the Seedance 2.0 series and raises the length of one generation from 15 to 30 seconds.
Ready to try Seedance 2.5?
The video tool opens with Seedance 2.5 already selected. The exact cost shows before every creation; you sign up when you generate.