AI Video, Image & Audio Models

A model is the generation engine and determines capabilities, quality, speed, and credit cost. A tool starts with a specific job—such as turning text into video or removing a background—and exposes the models and controls that fit that job. Open Tools when the job is already clear.

Start with the input you have (text, an owned still, or a reference pack), then duration and resolution. FLUX 3 covers a 5–20s clip from text or one still at 720p or 1080p. Seedance 2.5 covers a 4–30s one-take at 480p–1080p with a large pack. PixVerse V6 covers 1–15s from text or one photo with optional audio, multi-shot, and first-to-last transitions at 360p–1080p. Kling 3.0 uses 5/10/15s with an optional end frame. MiniMax H3 is for 4–15s instruction-led edits at 768p or 2K. Seedance 2.0 stays on 4–15s at 720p–4K. Veo 3.1 and Veo 3.1 Fast stay on 4/6/8s, and Gemini Omni Flash stays at 720p for 3–10s drafts. Exact credits are shown before you generate.

Use GPT Image 2.5 for Flare or Sunburst generate-and-edit at 1K–4K, GPT Image 2 for campaign stills and precise edits up to 4K, Seedream 5.0 Pro for dense multilingual layouts at 1K–2K, or Nano Banana 2 for faster prompt-faithful stills and edits at 1K–4K. Compare live credit ranges in the catalog before you open Create.

Use Eleven v3 for a single-speaker, emotion-tagged voiceover from a script. Use Seed Audio 1.0 when the brief is a timed sound scene with dialogue, ambience, or score cues. Both are available when your balance covers the displayed credit quote.

Every currently available model can run on the Free plan when you have enough credits. Free delivery limits, watermarking, queue priority, and evaluation-only terms still apply. See Pricing for plan details.