Seedance 2.5 US
ByteDance's US-hosted video model — 30-second single takes with synced audio and up to 50 multimodal references.
curl https://api.venice.ai/api/v1/video/queue \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "seedance-2-5-us-text-to-video-private",
"prompt": "Aerial drone shot over a misty mountain valley at golden hour",
"duration": "5s",
"resolution": "720p",
"aspect_ratio": "16:9"
}'
# Use the returned queue_id with https://api.venice.ai/api/v1/video/retrieve.
# Call /video/complete after downloading if needed.Overview
What is Seedance 2.5 US
Seedance 2.5 US is the US-hosted edition of ByteDance's Seedance 2.5 video generation model. It creates clips of 4 to 30 seconds at 480p, 720p, or 1080p with native synchronized audio, working from a text prompt, from a still image, or from up to 50 multimodal references including images, video clips, and audio.
Running it privately on Venice
On Venice, Seedance 2.5 US runs under the private tier with zero retention — your prompts and reference assets are never stored, profiled, or used for training. You get the same 30-second, audio-synced generations ByteDance sells elsewhere, without feeding a Big-Tech data pipeline. Pay per clip from $1.16 instead of a subscription, and switch between the text-to-video, image-to-video, and reference-to-video variants freely.
Specifications
Datasheet
- Maker
- ByteDance (Seed team)
- Modality
- Text-to-video, image-to-video, reference-to-video — all with native audio
- License
- Proprietary
- Clip lengths
- 4s – 30s
- Resolutions
- 480p, 720p, 1080p
- Mode
- text-to-video
- Audio
- Yes
- Prompt limit
- 15,000 chars
- References
- Up to 30 images, 10 video clips, and 10 audio clips per generation
- Reference video
- Accepted (up to 10 clips) — maximum duration not published by ByteDance
- Released
- July 31, 2026 (Seedance 2.5); on Venice since September 2026
- Aspect ratios
- auto, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16
- Privacy on Venice
- Private — zero retention
- Available on Venice since
- Sep 2026
Assessment
Strengths and limitations
- Longest single-pass clips of any video model on Venice — 30 seconds with synchronized audio generated in the same pass, double the 15-second ceiling of Grok Imagine 1.5, Wan 2.7 Enhanced, and MiniMax H3 Max.
- Deep multimodal reference control: up to 30 images, 10 video clips, and 10 audio clips in one generation, covering motion, clay-render, and creative reference styles.
- First and last frame control, so you can bracket a shot and let the model fill the arc between two stills.
- Multi-round extension: a 30-second clip can be extended in follow-up rounds to build multi-minute content with a consistent audiovisual language.
- Full 15,000-character prompt budget supports shot-by-shot scripting with camera movement and performance blocking.
- Three entry modes on one family: text-to-video, image-to-video, and reference-to-video for keeping a character or product consistent across shots.
- Closed and proprietary: no open weights, so it cannot be self-hosted or fine-tuned. Wan 2.7 and Wan 2.7 Enhanced are the open-weights alternatives on Venice.
- Stops at 1080p: there is no 4K option on this model.
- On Venice it runs in the private zero-retention tier, but not inside a TEE and not end-to-end encrypted — privacy comes from Venice not storing prompts, not from cryptographic isolation.
- The maximum duration of an accepted reference video is not published, so heavy video-reference workflows may need trial and error.
Use cases
What it is good for
- 01Long-form single-take ads and product stories that need 20–30 seconds with sound, not stitched 5-second fragments.
- 02Character- or product-consistent campaigns using the reference-to-video variant across multiple clips.
- 03Animating brand stills into motion with the image-to-video variant, bracketed by first and last frames.
- 04Vertical social content at 9:16 and cinematic trailers at 21:9 from one model.
- 05Multi-minute narrative pieces assembled by extending 30-second generations in rounds.
Prompting
Getting better results
Use the 15,000-character budget to write shot-by-shot direction — camera moves, blocking, and scene changes — rather than one-line descriptions.
Iterate at 480p ($1.16/4s clip), then re-render the winning prompt at 1080p once the motion is right.
For image-to-video, supply both a first and a last frame so the model interpolates a controlled arc instead of drifting.
Pass a reference image of your character or product to hold identity consistent, and add an audio reference when the soundtrack must match.
Pick 21:9 for cinematic takes and 9:16 for Reels or TikTok — set the aspect ratio explicitly instead of leaving it on auto.
Generate a 30-second base clip, then use extension rounds to continue the story rather than prompting a fresh unrelated clip.
Samples
Sample outputs
Generated on Venice with our standard prompt suite — the same prompts we run through every model of this type, so you can judge it like-for-like.
Calm ocean waves rolling onto a black-sand beach at sunrise, golden light on wet sand, foam dissolving into the shore, a single silhouetted figure at the waterline, smooth continuous forward push, serene cinematic atmosphere
A slow tracking shot through a rain-soaked Tokyo street at night, neon reflecting in puddles, steam rising from a food cart, a person with a translucent umbrella, shallow depth of field, teal-and-orange grade, smooth steady camera
Alternatives
How it compares
| Model | Best for | Max duration | Max resolution | Native audio |
|---|---|---|---|---|
| Seedance 2.5 US | Long single takes & multimodal references | 30s | 1080p | Yes |
| Grok Imagine 1.5 | Fast stylized clips | 15s | 1080p | Yes |
| Wan 2.7 Enhanced | Open-weights video with audio | 15s | 1080p | Yes |
| MiniMax H3 Max | High-motion short clips | 15s | 1080P | Yes |
| Vidu Q3 | Character-consistent shorts | 16s | 1080p | Yes |
Seedance 2.5 US is the right pick when the brief is a continuous 30-second take with synced audio and tight control over characters, products, or style via references. If open weights or sub-15-second speed matter more, look at Wan 2.7 Enhanced or Grok Imagine 1.5.
API
Call it from your code
Venice exposes this model through the REST API. Queue a generation with the model id.
curl https://api.venice.ai/api/v1/video/queue \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "seedance-2-5-us-text-to-video-private",
"prompt": "Aerial drone shot over a misty mountain valley at golden hour",
"duration": "5s",
"resolution": "720p",
"aspect_ratio": "16:9"
}'
# Use the returned queue_id with https://api.venice.ai/api/v1/video/retrieve.
# Call /video/complete after downloading if needed.Pricing
What it costs on Venice
Pay per clip on Venice — price scales with resolution and duration (4s–30s), from $1.16.
New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.
FAQ
Frequently asked questions
Seedance 2.5 US is the US-hosted edition of ByteDance's Seedance 2.5 video generation model, released July 31, 2026. It generates 4-to-30-second clips at 480p, 720p, or 1080p with native synchronized audio, from text prompts, still images, or multimodal references.
You pay per clip, with the price scaling by resolution and duration. A 4-second clip starts at $1.16 at 480p, $2.50 at 720p, and $6.14 at 1080p, with clips running up to 30 seconds. There is no subscription requirement.
It is neither free-software nor open-source — Seedance 2.5 US is proprietary to ByteDance with no published weights, so it cannot be self-hosted or fine-tuned. On Venice you can try it with free welcome credits before paying per clip. If open weights are a requirement, Wan 2.7 and Wan 2.7 Enhanced are the open alternatives on Venice.
Yes. The image-to-video variant (model id seedance-2-5-us-image-to-video-private) animates a still image into motion, with the same 4-to-30-second clip lengths, 480p/720p/1080p resolutions, and native audio as the text-to-video variant. A reference-to-video variant (seedance-2-5-us-reference-to-video-private) is also available for keeping a subject consistent.
Up to 30 images, 10 video clips, and 10 audio clips in a single generation, covering motion, clay-render, and creative reference styles. It also accepts first and last frame images to bracket a shot. ByteDance has not published the maximum duration of an accepted reference video.
Venice runs Seedance 2.5 US under its private tier with zero retention — prompts and reference assets are not stored, profiled, or used for training. Note that it is not end-to-end encrypted or TEE-isolated; the privacy guarantee is Venice's no-storage policy, available in the app or via the Venice API.
Seedance 2.5 US wins on length and control: 30-second single takes with synced audio versus Grok Imagine 1.5's 15-second ceiling, plus a far deeper reference system (up to 50 assets). Grok Imagine 1.5 is the simpler choice for quick stylized short clips where 15 seconds is enough.
A single generation runs from 4 to 30 seconds at any supported resolution, and the model supports multiple rounds of extension, so creators can chain 30-second takes into multi-minute content with a consistent audiovisual language.
Run Seedance 2.5 US privately
No prompt logging. No data used for training.