-900x506.webp)
Seedance 2.5 AI Video Generator by ByteDance
Generate videos up to 30 seconds from text or images, with native audio and reference controls available directly in getimg.ai.
Seedance 2.5 is ByteDance's latest Seedance video model, built for longer audio-video generation and reference-guided storytelling. Output runs at 480p, 720p or 1080p across six aspect ratios.
Use Seedance 2.5 Directly in getimg.ai
Seedance 2.5 sits in the Create video generator the moment you open getimg.ai. No API keys, no separate provider account, no minimum spend to unlock the model.
Text and image to video
Write a prompt, or start from an image and describe what should happen next. Both routes run through the same Create video generator, with prompt enhancement available if you want a rough description tightened before it reaches the model.
First Frame, Last Frame, Reference Images
Anchor the opening frame, set where the clip lands, or guide people, objects and places with reference images. These are separate guidance modes, so you pick the one that fits the shot.
Native audio, six aspect ratios, up to 30 seconds
Audio is generated in the same pass as the picture. Choose 16:9, 9:16, 1:1, 4:3, 3:4 or 21:9, set a length between 4 and 30 seconds, and output at 480p, 720p or 1080p.
Model Capability vs Available on getimg.ai
ByteDance ships Seedance 2.5 with a wider feature set than any single platform exposes. These six controls are live in the getimg.ai Create video generator today.
Not currently available in the getimg.ai workflow: video references, audio references, timestamp editing, multi-round extension and green-screen editing. If you have seen those demonstrated for Seedance 2.5 elsewhere, they are model capabilities rather than getimg.ai controls.
Choose Your Image Guidance Method
Seedance 2.5 accepts image input three different ways in getimg.ai. They solve different problems, so it is worth knowing which one you actually need before you upload anything.
Reference Images and First/Last Frame are separate modes and cannot currently be combined in the same getimg.ai generation.
Video and audio references
Point at a clip, not just a still
Seedance 2.5 reads video files and audio files as references, alongside the images you already attach. Hand it a clip to follow for motion or a sound file to work against, and it studies both while it generates.
You can bring in up to three video references and three audio references in a single generation, and mix them with image references in the same clip. Seedance 2.5 also takes the longest attachments of the models that accept this: 30 seconds of video or audio, where the others stop at 15, so a full-length music track goes in without trimming.
How to use Seedance 2.5 on getimg.ai
Generating your first Seedance 2.5 clip takes a moment.
1. Open the Create video generator
Click here and select Seedance 2.5 from the model picker in the prompt box. Seedance 2.0, 2.0 Fast and 2.0 Mini sit in the same list if you need 4K or a lighter, lower-cost run.
2. Write a prompt
Describe the subject, action, environment, camera behavior and audio. See our video prompting guide and the prompt formula further down this page. Optionally upload an image as the First Frame, the Last Frame, or a reference.
3. Generate
Choose from six aspect ratios (16:9, 9:16, 1:1, 4:3, 3:4 or 21:9), pick a length between 4 and 30 seconds, and set output to 480p, 720p or 1080p. Video and audio are generated together in a single pass.
Cinematic continuous long-take action shot. Realistic historical style, 16:9. Audio: Intense, accelerating tribal drum beats, muddy footsteps, and muffled battle cries. Extreme low-angle tracking shot moving backward rapidly. A fierce Viking warrior sprints through a dense, foggy pine forest during heavy rain. He parries a spear thrust without breaking stride, and vaults over a fallen wet tree trunk. The camera lifts over the trunk with him, maintaining perfect focus. He lands heavily, sliding slightly in the thick mud, displaying realistic weight and momentum, then immediately throws a hand-axe directly past the camera lens. Muted color palette, highly realistic physics, cinematic motion blur, raw and gritty atmosphere.
See What Seedance 2.5 Does with 30 Seconds
This is one generation, not a stitched edit: a single 30-second Seedance 2.5 clip at 720p, produced from one timeline prompt. The cutting, the on-screen title cards, the score and a brand name that stays consistent from the opening card through a product close-up to the end frame all came out of the same pass. Sound on.
Length changes what a clip can be, which is the practical reason to care about 30 seconds.
Longer Video. Native Audio. Reference Control.
Seedance 2.5 improves motion stability, multi-subject scenes, shot transitions and longer-form continuity compared with earlier Seedance generations. Complex interactions can still require multiple generations or additional editing.
Up to 30 seconds, text or image input, audio included in the same generation pass.
Cinematic macro nature ASMR, 16:9. No music, pure environmental sounds. Shot 1: Extreme close-up of a dark, damp mossy forest floor. A glowing amethyst-colored seed rests on the moss. Crisp sounds of water droplets hitting the leaves. Shot 2: Time-lapse effect combined with real-time fluid camera motion. The seed cracks open with a sharp, crisp crystalline "tink." A translucent, glowing purple stem rapidly grows upward. Tiny specks of bioluminescent pollen drift into the air, illuminating the dark background with soft, bokeh halos. Focus strictly on the hyper-detailed textures of the crystal petals, the refraction of light through the glass-like plant, and the moist, magical atmosphere.
Text to video that follows your direction
Describe the subject, action, environment, camera behavior, pacing and audio. Seedance 2.5 uses those instructions to guide shot structure and movement across the generated clip.
Multi-subject scenes and implied camera work hold together across the clip more reliably than in earlier Seedance versions, though busy interactions may still need a second pass.
16:9, hyper-realistic cinematic street racing shot. Audio: High-pitched engine revving, aggressive tire screech, and rain hitting metal. Camera starts low to the ground on a wet asphalt hairpin curve at night. A matte-black vintage Porsche 911 drifts aggressively into frame. The camera executes a fast whip-pan to the right, perfectly tracking the car's speed and keeping the car tack-sharp. The car slides out of frame. The camera abruptly stops panning and immediately rack focuses from the distant car taillights to a wet, crushed soda can resting on the asphalt in the extreme foreground. Perfect water physics as tires kick up rain.
How to Prompt Seedance 2.5
A reliable Seedance 2.5 prompt follows one order: scene, subject, action, camera, timing, lighting, pacing, audio. You do not need every element every time, but that sequence keeps the model from guessing.
Cinematic rainy-night street scene. A woman in a yellow coat runs toward a waiting taxi. Low tracking shot moving backward in front of her, slight handheld movement, neon reflections on wet pavement. At 8 seconds she reaches the car and opens the door. Distant traffic, rain, footsteps, no music.
Naming the audio matters as much as naming the shot. If you do not describe the sound, the model decides it for you.
Premium product commercial, 16:9, clean studio lighting, glossy dark stone counter, shallow depth of field. 0-5s: A matte black espresso machine sits in soft rim light against a dark backdrop. Slow orbit from left to right as steam begins to rise from the group head. 5-10s: Close-up as a double shot pours into a clear glass cup, crema forming and layering in slow motion. The camera pushes in on the pour. 10-15s: Pull back to a wide shot. The finished cup rests on the counter as warm morning light widens across the scene and the steam drifts upward. Audio: low grinder hum, the hiss of the pump, liquid pouring into glass, soft ambient room tone. No music, no voiceover.
Prompting a 30-Second Clip
At 30 seconds, timing carries more weight than description. Give the model a timeline and it has something to structure the clip around, which is how the 30-second example above was prompted.
Write each segment as its own beat: what happens, how the camera behaves, and what the audio does. Four beats is usually enough.
0 to 8s: setup
Establish the place, the subject and the mood. Keep the camera move simple so the viewer can read the scene before anything happens. Name the ambient sound here too, because it sets the floor for everything that follows.
Audio generated with the video. Not after it.
A generated soundtrack can combine dialogue, ambience, sound effects and music in the same audio-video generation. You describe the sound you want in the same prompt as the picture.
Audio and video are generated jointly and respond to the same prompt and scene structure. The output is a single mixed track, not separate stems to assemble afterwards.
Highly realistic sci-fi drama, 16:9. Single continuous cinematic shot. Audio features the rhythmic, isolated sound of heavy breathing inside a spacesuit helmet, backed by a subtle, eerie electronic drone and faint radio static. An astronaut stands frozen in a massive, dark alien cavern illuminated only by the harsh white beam of their shoulder lamp. The camera executes a slow, tense dolly-in toward the astronaut's reflective gold visor. A trembling, awestruck voice comes over the comms channel: "Houston, you're not going to believe this. The scans were wrong. These are not rock formations, they're hives, and they're hatching." The visor reflection reveals the beam of light sweeping across a massive, translucent egg pulsing with a deep bioluminescent green heartbeat, starting to hatch.
Seedance 2.5 or Seedance 2.0?
The Seedance family splits on a real trade-off rather than a straight quality ranking. Pick by the constraint that matters most for the shot.
Need the longest clip? Seedance 2.5
Up to 30 seconds in a single generation, at 480p, 720p or 1080p, across six aspect ratios. The pick for longer storytelling and longer continuous action.
Need 4K? Seedance 2.0
Up to 15 seconds, with 1080p and 4K output and the same six aspect ratios. Choose this when native output resolution matters more than maximum length.
Need a lighter run? 2.0 Fast or Mini
Both generate at 720p and cost materially less per second. Fast keeps all six aspect ratios up to 15 seconds. Mini covers 16:9, 1:1 and 9:16 up to 12 seconds.
We Tested the Seedance Family on the Same Prompts
We ran identical prompts through Seedance 2.5, 2.0, 2.0 Fast and 2.0 Mini in getimg.ai and kept every output. The clips below are unedited, generated at the same length and aspect ratio with native audio on. Play them side by side and judge motion, prompt following and audio for yourself.
Top left: Seedance 2.5. Top right: Seedance 2.0. Bottom left: Seedance 2.0 Fast. Bottom right: Seedance 2.0 Mini.
Specs below are measured from the getimg.ai Create video generator rather than taken from provider marketing. Test 1, human action: a parkour runner vaulting a rooftop ledge at sunset. Five seconds, 16:9, audio on.
We ran three prompts rather than four. A long-narrative test is not a like-for-like comparison, because only Seedance 2.5 reaches 30 seconds: 2.0 and 2.0 Fast stop at 15, and Mini at 12. That ceiling difference is shown in the 30-second example above.
Seedance 2.5
Up to 30s. 480p, 720p, 1080p. Six aspect ratios. First Frame, Last Frame, Reference Images. Native audio. 240 credits per second at 720p, 540 at 1080p. Best for longer storytelling and continuous action.
Seedance 2.0
Up to 15s. Up to 4K. Six aspect ratios. First Frame, Last Frame, Reference Images. Native audio. 350 credits per second at 1080p, 1,400 at 4K. Best when output resolution is the priority.
Seedance 2.0 Fast
Up to 15s. 480p and 720p. Six aspect ratios. First Frame, Last Frame, Reference Images. Native audio. 120 credits per second at 720p. Best for volume work at 720p.
Seedance 2.0 Mini
Up to 12s. 480p and 720p. 16:9, 1:1 and 9:16. First Frame, Last Frame, Reference Images. Native audio. 75 credits per second at 720p. The lightest option in the family.
Test 2: Product commercial
A pair of white leather sneakers held still on a polished concrete pedestal while the camera orbits around them, soft overhead light sweeping across the leather. Five seconds, 16:9, audio on.
The prompt asks the camera to move and the product to stay put, which is the harder of the two readings. Watch whether each model orbits the shot or quietly spins the shoes instead, and how product geometry holds up as the angle changes.
Top left: Seedance 2.5. Top right: Seedance 2.0. Bottom left: Seedance 2.0 Fast. Bottom right: Seedance 2.0 Mini.
Test 3: Dialogue
Two colleagues at a cafe table in morning light, one line of spoken dialogue, quiet ambience, no music. Five seconds, 16:9, audio on.
Dialogue is the hardest of the three tests. Listen for voice timing against lip movement, and for how each model balances speech against the room.
Top left: Seedance 2.5. Top right: Seedance 2.0. Bottom left: Seedance 2.0 Fast. Bottom right: Seedance 2.0 Mini.
Frequently Asked Questions
Seedance 2.5 is ByteDance's latest Seedance video model, developed by ByteDance Seed. It generates video clips from a text prompt (Text to Video) or from an uploaded image (Image to Video), with audio generated in the same pass as the picture.
In getimg.ai, Seedance 2.5 produces single clips between 4 and 30 seconds at 480p, 720p or 1080p, across six aspect ratios, with First Frame, Last Frame and Reference Image guidance.
Open the Video Generator on getimg.ai and start a Create video generation. Pick Seedance 2.5 from the model picker in the prompt box. Seedance 2.0, 2.0 Fast, 2.0 Mini, 1.5 Pro and 1.0 Pro Fast sit in the same list.
From text: write the prompt, set a length between 4 and 30 seconds, choose an aspect ratio and output quality, then generate. From an image: upload it as the First Frame, describe what should happen, and run it. Everything runs in the browser, with nothing to install.
Clear, specific prompts work better than long ones. Name the subject, the action, the environment, the camera behavior and the audio you want.
Yes. Seedance 2.5 is one of the models behind the getimg.ai AI video editor: add a clip as a video reference in Create video, describe the change, like a new background, a different outfit or a new visual style, and get a new version that closely follows the original movement and timing. You can attach up to three video references of up to 30 seconds each, together with audio tracks and images, and your original stays as it was.
Not in the same generation. Reference Images and First/Last Frame are separate guidance modes in the current getimg.ai workflow, so you choose one per clip.
Use First Frame, or First + Last Frame, when a specific frame has to be exact. Use Reference Images when you want to guide people, objects, places or overall visual direction without fixing any one frame.
All three generate video with native audio, but they trade off length, resolution and control.
Seedance 2.5 reaches 30 seconds in a single clip and outputs at 480p, 720p or 1080p. Seedance 2.0 stops at 15 seconds but outputs up to 4K. Both cover six aspect ratios, including 21:9 and 4:3, and both accept First Frame, Last Frame and Reference Images. Seedance 1.5 Pro is the lighter short-form option: three aspect ratios (1:1, 16:9, 9:16), First Frame and Last Frame control, and clips up to 12 seconds.
Reach for Seedance 2.5 when you need one long continuous take. Choose Seedance 2.0 when resolution is the priority. Seedance 1.5 Pro stays a solid, lower-cost pick for standard short clips.
Seedance 2.5 uses a multimodal architecture in which audio and video are synthesized in a single generation pass. Text and image inputs inform composition, motion, camera movement and audio together rather than in sequence.
Because the audio is generated from the same prompt and scene structure as the picture, it corresponds to what is on screen instead of being sourced separately and layered on afterwards.
Not currently in our public API catalog. Seedance 2.5 is available in the getimg.ai web app.
Our API currently includes selected earlier Seedance model: the Seedance 1.5 Pro API.
API access runs on the same account but bills separately on a top-up basis rather than subscription credits, so you can generate programmatically and scale at your own pace.
Paid plans include commercial usage rights for eligible generated content under our terms. Uploaded references, third-party intellectual property and model-specific restrictions may still apply.
Generate your first 30-second Seedance 2.5 clip
Seedance 2.5 is in the Create video generator on getimg.ai, with native audio, First Frame, Last Frame and reference controls. Paid plans include commercial usage rights for eligible generated content under our terms.