Create something worth keeping

Doubao Seedance 2.0 is ByteDance's newest video generation model. It excels at multi-shot storytelling with coherent characters, handles fast action and camera movement gracefully, and supports longer clips — up to around 15 seconds in one generation.

Doubao Seedance 2.0 Fast is the speed-optimized variant of Seedance 2.0. Same prompt language and visual style, delivered quicker and at a lower credit cost — ideal for drafting shots, exploring ideas, and high-volume social content.

Seedance 1.5 Pro is ByteDance's proven workhorse for AI video. It offers dependable quality for both text-to-video and image-to-video, with stable motion and good scene consistency — a safe default when you want predictable results.

Veo 3.1 is Google DeepMind's flagship video generation model, widely regarded as the quality benchmark for AI video. It produces cinematic footage with film-like lighting, coherent physics, and native audio generated together with the picture.

Veo 3.1 Fast is the quicker, lower-cost variant of Google's Veo 3.1. You keep the Veo look and prompt behavior while paying roughly 40% fewer credits per clip — the sweet spot for social video and rapid iteration.

Create video from a prompt with Seedance 2.5. Credits depend on output duration and resolution.

Create video from a prompt with MiniMax H3. Credits depend on output duration and resolution.

Create video from a prompt with Wan 3.0. Credits depend on output duration and resolution.

Create video from a prompt with Wan 3.0 Prime. Credits depend on output duration and resolution.

Nano Banana Pro is Google's top-tier image generation and editing model. It is known for rendering readable text inside images, keeping characters consistent across edits, and following complex instructions precisely — a favorite for posters, ads, and product shots.

Nano Banana 2 is the latest mainline release of Google's popular image model. It balances speed and quality for everyday creation — social graphics, thumbnails, concept art — and handles both generation from scratch and edits of an uploaded picture.

Seedream 5.0 is ByteDance's flagship image model and the best value in the catalog at 4 credits per image. It produces high-resolution, aesthetically polished results and is particularly strong at photorealistic scenes and stylized posters.

GPT Image 2 is OpenAI's image generation model, built on deep world knowledge and strong prompt understanding. It shines when a prompt needs reasoning — accurate diagrams, brand-consistent designs, and scenes with many precise details.

Qwen Image is Alibaba's image generation model, best known for rendering Chinese and English text beautifully inside images. At 7 credits it is a cost-effective choice for bilingual posters, e-commerce banners, and social media graphics.

Suno Music turns a plain-language idea into a complete song. Describe the genre, mood, instruments, vocal direction, and story you want, and the model builds a structured track rather than a short sound effect. It works well for demos, social videos, podcast themes, campaign concepts, and rapid songwriting experiments. MuseFable sends the request through AIAPI as an asynchronous music task and keeps the finished audio in your generation history. AIAPI currently marks the Suno route experimental, so generation time and upstream availability can vary.

MiniMax Music uses the Music 2.6 generation model to turn a prompt into a complete audio track. You can describe genre, mood, arrangement, vocal character, and lyrical theme, or request instrumental music for a specific scene. It is a practical choice for song demos, branded background music, short-form video, podcasts, games, and creative ideation. MuseFable submits the request through AIAPI's dedicated music category, tracks the asynchronous task, and exposes the finished audio for preview, download, and project reuse without requiring local audio software.

MiniMax Speech 2.8 Turbo is a state-of-the-art text-to-speech model that turns scripts into lifelike voiceovers in seconds. It supports multiple languages and speaking styles, making it a natural fit for video narration, ads, and audiobook drafts.

Claude Sonnet 4.6 is Anthropic's writing-focused model, prized for natural tone and nuanced instruction-following. Use it to draft video scripts, ad copy, captions, and prompt ideas that you can feed straight into the image and video models here.

Gemini 3 Pro is Google's flagship reasoning model. It handles long, complex briefs — research summaries, content plans, multi-scene scripts — and pairs naturally with Veo and Nano Banana when you plan a full content pack in one place.

DeepSeek V3.2 is a highly cost-efficient open model with strong bilingual writing ability. It is a practical choice for everyday copy — titles, descriptions, translations, and prompt drafting — especially when Chinese output quality matters.