Modèles Bytedance

Seedance 2.0 est le modèle de fondation vidéo multimodal de nouvelle génération de ByteDance, axé sur la génération de vidéos narratives cinématographiques à plans multiples. À l’inverse des démonstrations texte‑vers‑vidéo à plan unique, Seedance 2.0 met l’accent sur un contrôle basé sur des références (images, clips courts, audio), la cohérence des personnages et du style d’un plan à l’autre, ainsi qu’une synchronisation audio/vidéo native — afin de rendre la vidéo IA utile aux flux de travail professionnels de création et de prévisualisation.

Seedance 2.5 is a next-generation audio-video joint generation model built for 30-second storytelling, with precise reference control and powerful editing capabilities.

Seedream 5.0 Lite is a unified multimodal image generation model endowed with deep thinking andonline search capabilities, featuring an all-round upgrade in its understanding, reasoning and generationcapabilities.

🔹 Doubao Seed 2.0 Series doubao-seed-2-0-code-preview-260215 Focuses on long-chain reasoning capabilities and complex task stability, adapted for complex scenarios in real business environments. As the coding-enhanced version of Seed 2.0, it is better suited for Agentic Coding. doubao-seed-2-0-lite-260215 Balances generation quality with response speed, making it suitable as a general-purpose production-grade model. doubao-seed-2-0-mini-260215 Designed for low-latency, high-concurrency, and cost-sensitive scenarios. It emphasizes rapid response and flexible inference deployment, supporting four-level thinking and multimodal understanding capabilities.

Doubao-Seed-1.8 is optimized for multimodal agent scenarios. In terms of agent capabilities, tool use and complex command compliance have been significantly enhanced. Regarding multimodal understanding, basic visual capabilities have been significantly improved, enabling low-frame-rate understanding of extremely long videos. Video motion understanding, complex spatial understanding, and document structure parsing capabilities have also been optimized, and intelligent context management is now natively supported, allowing users to configure context strategies.

Seedream 4.5 est le modèle d’images multimodal de ByteDance/Seed (texte→image + édition d’images) qui met l’accent sur une fidélité d’image de niveau production, un respect renforcé des prompts et une cohérence d’édition nettement améliorée (préservation du sujet, rendu du texte/de la typographie et réalisme des visages).

Seedance is ByteDance’s next-generation multimodal video foundation model focused on cinematic, multi-shot narrative video generation. Unlike single-shot text-to-video demos, Seedance 2.0 emphasizes reference-based control (images, short clips, audio), coherent character/style consistency across shots, and native audio/video synchronization — aiming to make AI video useful for professional creative and previsualization workflows.

Seedream 5.0 Pro is designed for high-precision image generation scenarios, providing more precise control over position and elements. It supports generating images from text, single images, and multiple reference images.

Découvrez l'API Doubao Seedream 4-0.

doubao-seed-evolving is designed for coding and agent scenarios, delivering weekly capability updates through a unified model ID.

doubao-seed-2-1-turbo-260628 balances performance and cost, with upgraded coding, agent, and multimodal capabilities for real-world complex tasks.

doubao-seed-2-1-pro-260628 is a next-generation production-ready model with upgraded coding, agent, and multimodal capabilities for complex enterprise tasks.