SeeDance 2.5 by ByteDance: 4–30s multimodal video, dense refs, no face restrictions on Ganabgen AI vs Jimeng/Dreamina. Policy applies.
SeeDance 2.5 is ByteDance's long-form multimodal video model on Ganabgen AI — text-to-video, required first/last frames, or reference-heavy generation with [ImageN], [VideoN], and [AudioN] placeholders in your prompt.
On Ganabgen AI, SeeDance 2.5 runs without the face restrictions ByteDance enforces on its official Jimeng and Dreamina apps — useful for talent-forward reference boards and portrait-heavy prompts. You still need rights to any likeness you generate; standard platform policy applies.
Generate 4–30 second clips at 480p or 720p with synchronized voice, sound effects, and music. Up to 30 reference images, 10 video clips, and 10 audio tracks per request (50 assets total). First/last frame mode requires both frames and uses adaptive aspect ratio.
At ~176 credits for a typical 5s 480p clip (dry-run pricing), SeeDance 2.5 targets finished narrative units and dense reference control. For 1080p omni-reference at lower cost, compare [SeeDance 2.0 Pro](/video/seedance-2-0).
Generate with SeeDance 2.5 by ByteDance on Picturesque. Browse all video models.
| Credits | 120 credits |
|---|---|
| Resolution | 480p, 720p |
| Duration | 4–30s |
| Aspect ratios | adaptive, 1:1, 3:4, 4:3, 9:16, 16:9, 21:9 |
| Max references | 30 |
| Audio | Native / synced audio |
SeeDance 2.5 on Picturesque runs without face restrictions — unlike ByteDance's Jimeng and Dreamina apps, where heavy face censorship followed the Hollywood launch demos. You still need rights to any likeness you generate; standard platform policy applies.
Yes. SeeDance 2.5 is live in Video Studio via bytedance/seedance-2.5 on our upstream provider. SeeDance 2.0 Pro, Fast, and Mini remain available for shorter 1080p workflows.
2.5 supports 4–30s clips (vs 4–15s), 30/10/10 image/video/audio references (vs 9 omni slots), and [ImageN]/[VideoN]/[AudioN] prompt tags. Output is 480p or 720p only on Picturesque.
About 176 credits for a 5-second 480p clip; 720p scales higher (~378 credits for 5s). Duration and resolution multiply linearly — dry-run in Video Studio is authoritative.
Text-to-video, first + last frame (both required; adaptive aspect), and multimodal reference with at least one image, video, or audio URL. Modes are mutually exclusive.
Upload refs in Ref Images mode and tag them in your prompt as [Image1], [Video1], [Audio1] in array order. Up to 30 images, 10 videos, and 10 audio clips (50 total).
Pick 2.5 for longer clips (up to 30s) and denser multimodal refs. Pick SeeDance 2.0 Pro for 1080p delivery and lower per-second cost on short hero shots.
Yes — generate_audio defaults on. Toggle synced voice, SFX, and music in the studio sound control.
Explore more on Picturesque: Home, AI Image Generator, AI Video Generator, AI Audio Studio, Edit & Upscale, Motion Control, Director, Cinematic ad skill, UGC ad skill, Guides & Tutorials, FAQ, Explore, Pricing, Referral program, MCP. Resources: Blog.
Legal & policies: Terms, Privacy, Refunds, Delivery, Content policy, Cookies.