Everything About Dola Seed
ByteDance's AI Models, Explained
Dola Seed is the overseas name for ByteDance's Seed model family — including the Seedance video models and the Dola-Seed-2.0-Preview general-purpose multimodal model. Seedance 2.5 launched July 31, 2026 with 30-second single takes and up to 50 reference inputs — but official access is still fragmented. We cut through the noise.
AI Video Generator
Powered by DolaSeed-2.5-Preview / Seedance 2.5 API
Be the first to know when our video generation feature launches.
At a Glance
#1 Video Model
Arena Ranking
6 multi-shot
Max Shots
30 seconds
Max Duration
Native + Lip-sync
Audio
Coming soon
Global API
What Is Dola Seed?
Dola Seed is the name the international community uses for ByteDance's Seed model family. It encompasses two major product lines: the Dola-Seed-2.0-Preview general-purpose multimodal model (text + vision), and the Seedance video generation models.
The official overseas model identifier is Dola-Seed-2.0-Preview, released on Arena.ai in February 2026 as a general-purpose multimodal model. Meanwhile, Seedance 2.0 — the video generation flagship — ranks #1 on the Artificial Analysis Video Arena leaderboard (Elo 1,269) ahead of Kling 3.0, Sora 2, and Veo 3. Seedance 2.5 followed on July 31, 2026: 30-second single-pass generation with multi-round extension to multi-minute videos, up to 50 multimodal references (30 images, 10 videos, 10 audio), and timestamp-level editing control.
Unlike many competitors that generate isolated clips, Seedance is architected for coherent multi-shot storytelling — generating up to 6 connected shots in a single prompt with consistent characters, camera motion, and narrative flow. It also supports a unique multi-modal reference system: attach up to 9 images, 3 video clips, and 3 audio files, then reference them in prompts using @image1, @video2, etc.
Core Capabilities
Generate up to 6 connected shots with a single prompt. AI maintains character consistency, narrative logic, and natural transitions across scenes.
2.0 accepts up to 12 files; 2.5 expands to 50 (30 images, 10 videos, 10 audio) plus clay-render 3D blockouts. Lock character faces, art styles, or camera angles.
Built-in sound effects, ambient textures, background music, and lip-synced dialogue in 8+ languages. On 2.5, voice, SFX, and music generate with the picture in a single pass.
Specify camera movements — pan, tilt, zoom, dolly, tracking — with precise prompt adherence.
Upload a portrait photo to generate video with lifelike facial expressions, full-body motion, and accurate lip sync. Added April 2026.
4–15 second clips on 2.0; 4–30 second single takes on 2.5, extendable to minutes. Multiple aspect ratios: 16:9, 9:16, 4:3, 3:4, 1:1, 21:9 ultrawide. Up to 720p via API; 2K on consumer platforms; native 4K on 2.5.
How to Access Dola Seed & Seedance
There is no official global product. Here's every verified access route, ranked by reliability.
| Route | Model | Best For | Access Difficulty | Cost | Reliability | Notes |
|---|---|---|---|---|---|---|
| MolyinEditor's Pick | 2.5 / 2.0 | Everyone | Easiest | Free credits + paid | Seedance 2.5 live since release day — 30s takes, up to 4K. No CN phone, no approval wait, no region lock. | |
| Jimeng (即梦) | 2.5 / 2.0 | Chinese users | Easy (CN phone) | Free credits + paid | Official consumer platform. Seedance 2.5 available at launch. Requires mainland China phone number. | |
| Doubao | 2.5 (Pro) | Mobile users | Medium (approval) | Free (10/day) | Seedance 2.5 available on Doubao Pro. Often requires Feishu group approval. Fixed 5s/10s duration. | |
| CapCut | 2.0 | Brazil, SEA, MX, etc. | Easy | Free limited-time | Rolling out in BR, ID, MY, MX, PH, TH, VN only. | |
| fal.ai | 2.0 | Developers | Easy | Pay-per-use | Official partner route. OpenAI-compatible API. No region lock. | |
| BytePlus ModelArk | 2.5 (soon) | Enterprise | Hard | $0.10–$0.80/min | Official enterprise API. Seedance 2.5 API announced with the July 31 launch, coming soon. Heavy onboarding. | |
| Dreamina (Global) | — | International | N/A | N/A | Unavailable | Does NOT have Seedance 2.0 yet. Do not subscribe expecting it. |
| US Direct | — | US users | N/A | N/A | Unavailable | Excluded from BytePlus rollout. Use fal.ai or wait. |
Recommended for...
- Creators who want Seedance 2.5 today → Molyin (2.5 live) or Jimeng
- Developers globally → fal.ai API
- Enterprise teams → BytePlus ModelArk (2.5 API announced, coming soon)
- US-based users → Molyin (no region lock) or fal.ai API
How Seedance Compares
| Model | Arena Elo | Native Audio | Multi-Shot | Max Duration | Global API | Best For |
|---|---|---|---|---|---|---|
| Seedance 2.5 | New — unranked | Yes | Yes | 30s | Announced | 30s one-takes, timestamped direction |
| Seedance 2.0 | 1,269 (#1) | Yes | Yes | 15s | Delayed | Character dialogue, multi-shot ads |
| Kling 3.0 | 1,248 (#2) | No | No | 120s | Yes | Long-form content, fast generation |
| Sora 2 | ~1,230 | 3rd party | No | 60s | Paused | Cinematic shorts, complex scenes |
| Veo 3.1 | ~1,220 | Ambient only | No | 8s | Yes | High-fidelity textures |
Source: Artificial Analysis Video Arena, March 2026. Elo ratings are dynamic and change as models are re-evaluated. Seedance 2.5 (released July 31, 2026) has not yet received an Arena rating.
What People Use Seedance For
Mini-series pilots, narrative content with consistent characters across scenes.
Hero shots, product reveals, lifestyle scenes with cinematic camera movement.
Audio-visual sync, mood-driven footage, abstract to cinematic sequences.
Storyboard-to-video for filmmakers, game trailer mockups, investor pitches.
Find the Right Prompt for Seedance
Browse curated Seedance prompts on AhaPrompt — copy one, tweak it, and paste it straight into Molyin to generate.
View Seedance PromptsFrom the 2.0 Copyright Dispute to Seedance 2.5
In February 2026, hours after Seedance 2.0's launch, AI-generated videos featuring Hollywood stars (Tom Cruise, Brad Pitt) and copyrighted characters (Superman, Wonder Woman, Joker) went viral. The Motion Picture Association (MPA) censured the model, and major studios — Disney, Warner Bros., Paramount Skydance, Netflix — issued cease-and-desist letters.
ByteDance had planned a global API rollout for February 24, 2026, but postponed it indefinitely to address these legal concerns. The situation has since evolved: Seedance 2.5 officially launched on July 31, 2026 on Jimeng and Doubao Pro, with API access via BytePlus ModelArk announced as coming soon.
What ByteDance has implemented since:
- Blocking recognizable real people as source material
- Copyright content filters
- C2PA metadata watermarking on all outputs
- A library of 10,000+ virtual people for approved commercial use