Gemini Omni

Capabilities

What it costs in credits

Derived from the same pricing source the app charges against, so these figures and the cost preview on the generation screen can never disagree.

Credits come from your plan and from credit packs. See the pricing page for the plan ladder.

When to use Gemini Omni

Gemini Omni Flash is Google's fast multimodal video model and the other place to go for talking content. Like HappyHorse it generates native audio and multilingual lip-sync on every clip, and it accepts text, an image and a large set of reference images. Unlike HappyHorse it will also take a finished clip back in and rework it: you hand it a video and a prompt, and it re-renders that footage the way you describe. Be precise about what that gives you, because the obvious assumption is that the clip gets longer, and that is the one thing this mode does not do. Nothing is added after the last frame of the video you supply. If the line you wrote runs longer than the take, the answer is a fresh generation or a different model, not this mode. What the mode is genuinely good for is a variation on a take you already like: change the setting behind the speaker, the wardrobe, the grade or the mood, and start from the framing and the delivery you already approved instead of rolling the dice again. Its duration menu is the shorter of the two talking models and it offers a single resolution, so the trade is control and speed against range. Reach for it when you want a spoken clip quickly, when you are working in a language other than English, or when you have a take that works and want a second version of it. It sits in the middle of the catalogue on price with audio included, which makes it reasonable value for talking content and poor value for anything silent, since you cannot switch the audio off. If the clip is going under a music bed or a separately recorded voiceover, generate it on Veo 3.1 Lite or Kling 2.6 Pro instead and keep the difference. If the piece needs to run past this model's duration ceiling, move to HappyHorse or Seedance, or build it out of several clips and cut them together. As with HappyHorse, remember that the dedicated talking-shot pipeline exists when you need the character to say exact words in the character's own voice.

Best for

• Short spoken beats and quick talking clips
• Multilingual social content
• A second version of a take whose delivery already works

Other models to consider