Seedance 3 vs Seedance 2.5: Release Date, Price, Prompts

Search for the Seedance 3 release date and you will find plenty of waitlists and very few facts. Here is what ByteDance has actually said, as of 29 September 2026, plus what Seedance 2.5 can do for your ads today, what it costs and how to prompt it.

Short answer: Seedance 3 has not been announced. ByteDance has published no name, date or feature list, so any release date you see online is a guess. The newest model is Seedance 2.5 (out since 31 July 2026): videos up to 30 seconds with native audio, up to 50 references and up to 1080p through the API.

Seedance 3: what is officially known?

Nothing yet. On the ByteDance Seed team's site, in the BytePlus API documentation and on Dreamina, Seedance 2.5 is the latest version. There is no announcement, no API model name and no official waitlist. "Seedance 3.5" does not officially exist either.

The only official look ahead sits at the end of the Seedance 2.5 launch post. The Seed team says it will keep working on more coherent storytelling, a more intuitive way to generate and edit, and a deeper grasp of real-world physics. It also names the weak spots of 2.5: complex motion does not always look natural, and scenes with several interacting subjects are less stable. A successor is the obvious next step, but its name and timing are unknown outside ByteDance.

The pace so far, according to Wikipedia: 1.0 in June 2025, 1.5 Pro in December 2025, 2.0 in February 2026 and 2.5 on 31 July 2026. That is a rhythm, not a promise.

Sites with "Seedance 3" in the domain and a sign-up box are not ByteDance channels; at least one says so itself at the bottom of its page. Do not pay for "early access".

What Seedance 2.5 can do for your ads today

ByteDance unveiled 2.5 in June 2026 at the Volcano Engine FORCE conference and released it on 31 July in its own apps. Since early August there has also been an API through BytePlus. What matters for advertisers:

• Length: 4 to 30 seconds in one generation, and you can extend a video. Dreamina also mentions a beta long-video mode of up to 180 seconds.
• References: up to 50 per request: 30 images, 10 videos and 10 audio clips (video and audio each capped at 30 seconds in total). Your product, location and brand look can all go in at once.
• Audio: speech, sound effects and music are generated with the picture. Speech works in 11 languages, including English and Spanish. Dutch, German and French are not on the list.
• Editing: change part of a clip, say seconds 4 to 6, without regenerating all of it.
• Formats: from 9:16 to 21:9, up to 1080p through the API. The launch coverage and Dreamina mention 4K, but the API documentation does not list 4K for 2.5.
• Draft mode: a cheap 480p preview to check the structure before you pay for the final render.
• Real faces: you cannot simply upload a photo of a real person as a reference; BytePlus offers separate routes such as preset digital characters (documentation).

Seedance 2.5 vs Seedance 2.0

Sources: the BytePlus tutorial, prompt guide and pricing page.

In short: 2.5 wins on long, coherent stories with many references. 2.0 still wins on maximum sharpness and price per second.

Seedance 2.5 price, and can you use it for free?

There are three routes:

1. The BytePlus API, if you build your own software. BytePlus bills per token; its own worked examples come to roughly $0.10 per second at 480p, $0.23 at 720p and $0.57 at 1080p without video input (pricing page). A 30-second 720p clip therefore costs around $7. Adding a reference video raises the price. Videos that fail, for instance on content moderation, are not billed.
2. Dreamina by CapCut. According to the Dreamina page, you get free daily credits to try Seedance 2.5. Paid plans started at €19 per month in late September 2026, with frequent promotions.
3. An ad platform such as Genly. There, Seedance 2.5 costs 2 credits per second, so 60 credits for 30 seconds. There is a 3-day free trial; plans are on the pricing page.

Seedance 2.5 prompt guide: the rules that matter for ads

ByteDance published its own prompt guide for 2.5. These are the rules that pay off most in advertising:

• Keep a fixed order: subject, action, setting, style, camera, sound.
• Give every reference a job. Number assets in upload order and state what to take from each: "@Image 1 is the product, keep the label exactly as shown."
• Use timestamps in whole seconds, like "0-3s ... 3-8s". Cram too much into one block and you get messy cuts.
• Separate sound with brackets: ( ) for music, < > for sound effects, { } for dialogue and 【】 for subtitles. For non-Chinese dialogue, name the language first.
• Add text and logos in the edit. The model misspells on-screen text, so leave space and add copy afterwards.
• Expect the odd exception. Even with "no subtitles" or "no BGM", subtitles or music occasionally appear. ByteDance says this has improved a lot but cannot be fully prevented.

Comparing models? See our Sora 2 prompt examples for ads.

Three Seedance 2.5 prompt examples for ads

All three use our demo brand Tropical Elixir, a bottle of tropical juice. Upload a product photo as @Image 1.

1. A 15-second product video

Vertical 9:16 product ad, 15 seconds, bright natural daylight, realistic commercial look. @Image 1 is the product: keep the bottle shape, label colours and cap exactly as in @Image 1.
0-3s: close-up, a hand places the bottle on a sunlit kitchen counter. <soft clink of glass on stone>
3-8s: slow push-in, the hand pours the juice into a glass with ice. <pouring, ice cracking>
8-12s: side view of a woman in her thirties taking a sip and smiling.
12-15s: static packshot of the bottle next to the glass, empty space above it.
(light acoustic guitar, upbeat, freshly composed, no vocals) No subtitles, no text on screen.

2. Faceless UGC, ready for your own voice-over

If your ad language is not one of the 11, generate the picture without dialogue and lay your own voice-over on top.

Vertical 9:16, 20 seconds, handheld smartphone footage, UGC style, natural window light in a small apartment. @Image 1 is the product.
0-4s: first-person view, hands take the bottle out of a paper delivery bag. <paper rustling>
4-10s: close-up, the hands turn the bottle toward the camera and shake it. <liquid sloshing>
10-16s: the hands pour a glass, then a hard cut to the glass on a desk next to a laptop.
16-20s: the hand raises the glass toward the camera.
Only natural room sound and action sounds. No dialogue, no music, no subtitles.

3. A 30-second story

Hook, problem, solution and packshot in one video, with no stitching.

Vertical 9:16 commercial, 30 seconds, realistic, warm colour grade. @Image 1 is the product. @Image 2 is the main character: keep her face and clothing consistent.
Shot 1 [0-4s]: the woman from @Image 2 yawns at her office desk, a cold coffee next to the keyboard. Hard cut.
Shot 2 [4-10s]: she takes the bottle from @Image 1 out of the office fridge. Close-up of the label.
Shot 3 [10-20s]: back at her desk, she drinks, straightens up and types with energy.
Shot 4 [20-26s]: evening, she walks out of the building smiling, bottle in hand.
Shot 5 [26-30s]: packshot of the bottle on a clean desk, empty space for text.
<quiet office ambience> (warm light electronic beat, freshly composed, no vocals). No subtitles, no text on screen.

Using Seedance 2.5 in Genly

Genly is a Dutch platform for creating Meta and TikTok ads. Seedance 2.5 is built into its video generator, into GenFlow and into the AI agent. What that solves:

• Dutch voice-over in the same workflow. In GenFlow you connect blocks for script, voice-over, image and video, and pick the model per block.
• Product and persona as references. A product photo can go in as the first frame or as a reference. Genly registers persona faces with the provider automatically, so they pass content review.
• Help with script and hook. The AI agent writes in your brand voice; you then finish the video in the editor.
• Honest limits. In Genly, 2.5 renders at 480p or 720p. For a sharper final render, choose Seedance 2.0.

Read more on the Seedance 2.5 in Genly page. New to AI video? Start with how to make AI video ads.

And Seedance 3? What you practise now carries over to the next model: strong references, a clear shot structure and a separate voice-over.

Frequently asked questions