HelloGen.ai
Seedance 2.5ByteDance’s newest video model, onlineSeedance 2.5 is the video model from ByteDance’s Seed team, announced at Volcano Engine FORCE on 23 June 2026 and opened to the public on 31 July: it holds a single take together for up to 30 seconds, generates its audio in the same latent space as the picture, follows prompts noticeably better than Seedance 2.0 and can redraw part of a frame without touching the rest. On HelloGen it renders 5- or 10-second clips at 480p or 720p from the Studio’s video composer, priced per second in credits.
A Pro-tier model, from 140 credits per clip. Pro subscribers go straight to the Studio; everyone else lands on the plans page first. A failed render is never charged.
Four upgrades over Seedance 2.0, and what they mean for the clips you make.
The headline claim is a native single-pass take of up to 30 seconds with no visible seams, and characters and environments that stay consistent from the first shot to the last. Even in a 10-second clip that shows up as steadier faces and fewer drifting details.
Dialogue, effects and ambience are generated in step with the motion rather than added afterwards, so a door slam lands on the frame the door closes. In the HelloGen composer, sound is a toggle that costs no extra credits.
The model accepts far more multimodal references than 2.0 — up to fifty at the API. In the composer you can attach up to nine reference images plus a reference video or audio track to lock a character, a style or a move.
ByteDance quotes roughly 20% better prompt adherence than Seedance 2.0, and a local re-draw ability that swaps a background or a product while the rest of the frame — motion, camera, lighting — stays put.
Exactly what the composer offers today — resolutions, clip lengths, sound and modes as you will find them.
| Plan | Pro |
|---|---|
| Resolution | 480p · 720p |
| Clip length | 5s / 10s |
| Sound | Optional — switch generated sound on or off before you submit |
| Modes | Text to video · Image to video · First and last frame · Reference guided |
| Reference inputs | Up to 9 reference images, plus one reference video or audio track |
| Aspect ratios | 21:9 · 16:9 · 4:3 · 1:1 · 3:4 · 9:16 |
| Credits | From 28 credits per second of output |
Priced per second of output, by resolution. The composer shows the exact figure before you submit, and a failed render is never charged.
| Resolution | Length | Credits | With a reference video |
|---|---|---|---|
| 480p | 5s | 140 | 85 |
| 480p | 10s | 280 | 170 |
| 720p | 5s | 315 | 190 |
| 720p | 10s | 630 | 380 |
With a reference video attached, both the reference seconds and the output seconds are billed, at the lower per-second rate shown. Pro includes 3,600 credits a month ($19.9/month, or $15.99/month billed yearly); credit packs start at $4.90 for 500 credits and never expire.
Open the Studio’s video tab and choose Seedance 2.5 from the model list — the button above preselects it. The credit price updates as you set resolution, length and sound.
Describe the scene shot by shot if you want a cut, or drop in a photo as the first frame. Add a last frame to control where the clip lands, or attach reference images and a clip to lock a character or a move.
Choose 480p or 720p, 5 or 10 seconds, switch generated sound on if you want it, then submit. Finished clips land in My Works; a failed render costs nothing.
Copy one into the composer and swap the nouns. Prompts are kept in English so they can be pasted as-is.
A short travel vlog in three shots: a wide establishing shot of a fishing village at dawn, a close-up of nets being hauled in, then a smiling fisherman holding up the catch. Golden light, gentle sea sounds, consistent handheld documentary look.
Bring this illustration to life: the fox in the drawing blinks, its ears twitch, then it trots out of frame to the right while leaves drift across the scene. Keep the watercolor texture exactly as it is.
Move from the empty desk in the first frame to the cluttered workspace in the last frame as a time-lapse: papers, mugs and a laptop appear one by one while daylight shifts to warm lamp light.
Using the reference images for the character and outfit, she skateboards down a sunlit boardwalk. Low-angle camera tracking alongside, wind in her hair, wheels rumbling on the boards, upbeat mood.
Pick Seedance 2.5 when a scene needs to survive a cut — several shots, a character who must look the same throughout, or a soundtrack that lands on the action. Seedance 2 is the choice for a 1080p frame; Seedance 2 Fast and 1.5 Pro cover quick, cheaper drafts on the Basic plan; MiniMax H3 is the alternative when you want 2K resolution and native sound at a lower per-second price.
| Model | Plan | Resolution | Length | Sound option | From |
|---|---|---|---|---|---|
| Seedance 2.5Long clips, up to 30s | Pro | 480p · 720p | 5s / 10s | ✓ | 140 credits |
| Seedance 2Flagship video quality | Pro | 480p · 720p · 1080p | 5s / 10s | — | 205 credits |
| Seedance 2 FastFast HD video | Basic | 480p · 720p | 5s / 10s | — | 165 credits |
| Seedance 1.5 ProFast, cinematic video | Basic | 480p · 720p · 1080p | 5s / 10s | ✓ | 35 credits |
 MiniMax H32K video with native sound | Pro | 720p · 2K | 5s / 10s | ✓ | 80 credits |
The credits column is each model’s cheapest combination; the composer prices the exact resolution, length and sound you pick.
Seedance 2.5 is a video generation model from ByteDance’s Seed team, announced at the Volcano Engine FORCE conference on 23 June 2026 and released publicly on 31 July 2026. It generates video with native audio from text, images and reference material, and its headline upgrade is consistency across a long single take — up to 30 seconds at the model level.
Four things: a native single-pass take of up to 30 seconds instead of stitched clips, audio generated in the same latent space as the picture, many more multimodal references per generation (up to fifty at the API, versus twelve), and a local re-draw ability that changes one part of a frame while the rest stays put. ByteDance also quotes roughly 20% better prompt adherence.
You choose 5 or 10 seconds per clip in the composer. The model itself can hold a take for up to 30 seconds; for a longer piece on HelloGen, use the last frame of one clip as the first frame of the next and join them in a video editor.
Yes. Seedance 2.5 generates its audio — dialogue, effects and ambience — together with the picture, so it stays in sync with the motion. In the HelloGen composer, sound is a toggle you switch on before submitting, and it costs no extra credits.
28 credits per second at 480p and 63 per second at 720p, so a 5-second 480p clip is 140 credits and a 10-second 720p clip is 630. When you attach a reference video the rate drops to 17 and 38 credits per second, billed on the reference seconds plus the output seconds. Pro includes 3,600 credits a month, and a failed render is never charged.
Seedance 2.5 is a Pro-tier model on HelloGen. Subscribe to Pro ($19.9 a month, or $15.99 a month billed yearly) and it appears in the video composer with 3,600 monthly credits; the button on this page takes subscribers straight to the Studio and everyone else to the plans page. Seedance 2 Fast and Seedance 1.5 Pro are the Basic-plan options in the same family.
Yes. In reference mode you can attach up to nine reference images plus one reference video or audio track, then describe the scene; the model uses them to keep a character, a style or a movement consistent. Reference videos of up to 15 seconds are accepted, and their length is included in the credit price.
Seedance 2.5 when the clip has to survive a cut or carry dialogue across a scene: multi-shot consistency is its strength. MiniMax H3 when you want a higher-resolution frame — native 2K — with sound at a lower per-second price. Both are Pro models, and both take text, a first frame, first-and-last frames or reference material.
Write the shots, switch sound on, and let Seedance 2.5 render them in one take.
✦ Try Seedance 2.5 nowMiniMax H3Native 2K at 24 fps with sound generated alongside the picture — text, image and reference modes, from 80 credits a clip.
All AI video models, side by side