Every model. One plan.

FLUX 3 Video for anything going public: sound with the picture, Full HD and up to 20 seconds, and FLUX 3 Video Draft to explore the same take at a third of the price. Veo 3.1 when realism and dialogue matter most and the shot fits in 8 seconds, with Veo 3.1 Fast and Lite as the cheaper ways to draft it. Seedance 2.5 for one continuous take up to 30 seconds, Seedance 2.0 for 4K or for steering a shot with reference clips and audio. Kling 3.0 is the cinematic all-rounder up to 15 seconds, Kling O3 Reference puts up to four people or products in one shot, and PixVerse V6 is the cheapest way to test a hook.

Klein 9B while you iterate, because it is two credits and instant. Flux 2 Pro for most finished work, and Flux 2 Max when the hero shot has to hold up at full size. GPT Image 2.5 Flare the moment there is text in the frame, and Sunburst when the frame will go through rounds of edits. FLUX 3 Image when one frame combines up to ten references or has to match a FLUX 3 Video. Nano Banana Pro for photoreal multi-image edits at 4K, Ideogram 3 and Recraft V4.1 for posters, logos and illustration, and Seedream for cheap 2K and 4K.

Yes. ElevenLabs v3 is the most expressive voiceover, Multilingual v2 narrates in 29 languages and Turbo v2.5 costs half as much. Chatterbox speaks in your own voice from a short sample. MiniMax Music 2.6 writes full songs with vocals, Eleven Music makes a track at the exact length you need and Lyria 3 sketches 30 second ideas. Eleven Sound Effects turns a description into a sound, ElevenLabs Dubbing translates a video's speech in the same voices, and VEED Clean Audio rescues a noisy recording. FLUX 3 Video, Veo 3.1, Kling, Wan and Seedance also generate sound with the picture.

Two ways. To make a photo talk, Kling Avatar 2 takes a portrait plus your recording, or a script read by an ElevenLabs voice or your cloned voice, and returns up to 60 seconds of talking video; the Pro version renders face and expression at higher fidelity. To put new words in a clip you already have, Kling Lip Sync handles clips up to 10 seconds and VEED Lip Sync goes to a minute. FLUX 3 Video can also speak a line you write, with matching lips, in a single render.

Free covers Flux 2 Klein 9B, FLUX 3 Image and the GPT Image models, for three watermarked images. Hobby adds Flux 2 Pro, Nano Banana, Seedream, Ideogram and Recraft, the Photo Relight and Photo Angles tools, and every voice, music and sound model. Starter adds video: FLUX 3 Video, Seedance, Veo, Kling, Wan, Hailuo and PixVerse, plus lip sync, motion, dubbing and the video tools. Flux 2 Max is Plus and up.

Each price on this page is the model's cheapest setting. Length, resolution, sound and quality raise it, and the exact cost of your settings is on the Generate button before you commit. Every model draws on the same pool of credits.

The FLUX, GPT Image, FLUX 3 Video and Seedance models are in the REST API and the MCP server. The others run in the web studio.

Never. New models land in the same studio, priced in the same credits, with no per model surcharge and no second subscription. Ultra gets them first.