How to Use Seedance 2.5: Step-by-Step Guide to 30-Second AI Videos

Seedance 2.5 is ByteDance's newest AI video model. It generates clips of up to 30 seconds in one pass with sound, follows up to 30 reference images, and — unlike Seedance 2.0 — can also edit and extend videos you already have. This guide walks through every way to use it, which settings to choose, and how to write prompts that get the shot you want on the first few tries.

You can follow along on SeeVid's AI Video Generator, which runs Seedance 2.5 in the browser with no API key or ByteDance account.

Pick the Right Mode First

Seedance 2.5 has five ways in. Choose the one that matches what you already have:

Step 1: Text-to-Video

1. Open the AI Video Generator, choose Text to Video and select Seedance 2.5.
2. Write your prompt — up to 10,000 characters, so there is room for detailed multi-shot scenes (see the formula below).
3. Choose your settings:
   • Aspect ratio: 16:9 for YouTube, 9:16 for TikTok and Reels, 1:1 for feeds, or Auto.
   • Duration: 4 to 30 seconds, or Auto to let the model choose a length that fits the prompt. Test with 5 seconds first.
   • Quality: 480p, 720p or 1080p.
   • Draft: turn on to preview cheaply (explained below).
   • Generate audio: on by default.
   • Web search: optional; lets the model look up current information for prompts about real-world things, such as a new product or today's weather. It is billed only if a search actually runs.
4. Click Generate Video. The credit cost is shown on the button before you start, and most clips are ready in a few minutes.

Step 2: Image-to-Video

Image-to-video keeps the look of your image and adds motion.

1. Choose Image to Video and select Seedance 2.5.
2. Upload one image as the first frame. Add a second image to set the last frame — the model creates the motion between the two.
3. Describe what happens, not what the picture looks like: "She turns toward the window and smiles as the camera slowly pushes in; curtains drift in the breeze."
4. The aspect ratio follows your image. Pick a duration and quality, then generate.

Step 3: Reference-to-Video (Up to 30 Images)

This is where Seedance 2.5 pulls ahead. Choose Reference to Video, select Seedance 2.5, and combine:

• Up to 30 images: characters, products, outfits, locations, style frames
• Up to 10 video clips (30 seconds in total): a camera move, a dance, an editing rhythm
• Up to 10 audio clips: music, a voice, a sound to follow

Each upload gets a tag — @image1, @video1, @audio1 — that you use in the prompt to say what it is for. Click the tag under an upload to insert it.

@image1 walks into the café from @image2 wearing the coat from @image3 and orders from @image4. Copy the slow handheld tracking shot from @video1. Cut to a close-up of the cup with the logo from @image5. Music: @audio1.

Be explicit about each reference's role ("the face from", "the camera move from", "the music from"). With many references, give every character and product a clear name in the prompt and keep using the same tag.

Note: reference images and audio are free, but reference videos are billed because the model processes them. The price on the Generate button already includes them.

Step 4: Edit an Existing Video

The AI Video Editor changes part of a clip while keeping its motion and camera.

1. Upload a clip between 4 and 30 seconds (up to 200 MB).
2. Optionally add reference images, for example the new product or outfit.
3. Describe the change: "Replace the red sneaker with the white sneaker from @image1. Keep everything else the same."
4. Pick the quality and generate. The output has the same length as your clip.

Good uses: swapping a product or packaging, changing a background or season, restyling a scene, fixing a detail without re-shooting.

Step 5: Extend a Video

The AI Video Extender continues a clip forward (what happens next) or backward (what happened before).

1. Upload the clip you want to extend.
2. Choose the direction and the length of the new part (4–30 seconds).
3. Describe the new action: "She opens the door and steps out into the rain."

Extending is the way to build a sequence longer than 30 seconds while keeping the same characters and look.

Draft Mode: Preview Cheaply, Then Upgrade the Same Take

In text-, image- and reference-to-video, the Draft switch renders a 480p preview at the 480p price. If you like the take, click Upgrade to 1080p within 24 hours and you get that exact video in 1080p — no regenerating and hoping for the same result. If you don't like it, you only paid for the draft.

Draft mode saves the most when you expect to try several takes before keeping one. See the Seedance 2.5 pricing guide for the numbers.

The Seedance 2.5 Prompt Formula

Strong prompts cover six things, roughly in this order:

1. Subject — who or what, with one or two defining details
2. Action — what happens, in sequence
3. Scene — where and when, lighting and weather
4. Camera — shot size and movement (tracking shot, dolly in, orbit, handheld, aerial)
5. Style — cinematic, documentary, anime, product commercial
6. Sound — dialogue in quotes, sound effects, music mood

Use the 30 Seconds: Write in Shots

A 30-second clip needs a plan. Split it into shots with rough timing:

Shot 1 (0–8 s): wide drone shot of a mountain road at sunrise, a red car climbing the curves. Shot 2 (8–16 s): inside the car, the driver smiles and turns up the radio. Shot 3 (16–24 s): tracking shot alongside the car through a pine forest. Shot 4 (24–30 s): the car stops at a viewpoint; the driver steps out and looks at the valley. Warm golden light, cinematic. Upbeat indie rock, engine hum, wind.

Dialogue

Put spoken lines in quotes and say who speaks: The old man looks up and says "You came back." Seedance 2.5 generates the voice and lip movement with the picture.

Settings Cheat Sheet

Common Mistakes (and Fixes)

• The character drifts in a long clip. Add a clear front-facing reference image, name the character, and tag it in every shot.
• Thirty seconds with one action. Long clips need a sequence of actions or shots; otherwise the motion stalls.
• References with unclear roles. "Use these images" leaves the model guessing; say what each tag is.
• An edit changes too much. Say what must stay the same: "Keep the person, camera and lighting unchanged."
• Close but not quite. Keep the prompt, change one detail, and generate again — each run is a new take.

FAQ

Is Seedance 2.5 free to use? New SeeVid users get free credits after verifying their email, enough to try Seedance 2.5. After that, generations use credits from a pack or subscription. See the Seedance 2.5 pricing guide.

How long does Seedance 2.5 take? Usually a few minutes. Longer clips, 1080p and reference videos take longer. You can leave the page; the result is saved in My Creations.

Can Seedance 2.5 make 4K video? Not on SeeVid at the moment — it goes up to 1080p. Use Seedance 2.0 for 4K, or upscale a 2.5 clip with the AI Video Upscaler.

Should I use Seedance 2.5 or 2.0? Use 2.5 for clips over 15 seconds, many references, editing and extending; use 2.0 for shorter, cheaper clips and 4K. Full comparison: Seedance 2.5 vs 2.0.

Can I use the videos commercially? You can use your generations in your own projects. Make sure you have the rights to everything you upload, and permission from anyone whose likeness appears.

Start Creating

Pick a mode, turn on Draft, and test your prompt with a 5-second clip. When the shot works, raise the length and quality — or upgrade the draft you already like.

Open Seedance 2.5 on SeeVid, and see how it stacks up against other models in Seedance 2.5 vs Kling 4.0 vs Veo 3.1 vs Wan 3.0.