It’s debatable whether Google’s Veo is the best AI video generator or has been eclipsed by something newer, but what’s not up for debate is its cost – it’s by far the most expensive option. Therefore, use these tips to minimize unusable generations.

For consistent characters in Flow, use “ingredients”. According to Google:

“An ingredient is a consistent visual element — a character, an object or a stylistic reference — that you can create from a text-to-image prompt with the help of Imagen or by uploading an image. You can add up to three ingredients per prompt by selecting “Ingredients to Video” and then generating or uploading the desired images.”

You should be able to add your two main characters this way, and keep them consistent with Ingredients to Video.

Another way to generate a new clip with the same character is to Jump To it. According to Google:

“Transition a character or object to a completely new setting while preserving their appearance from the previous shot. It’s like teleporting your subject, saving you from recreating them for a new scene.”

In general, you’re going to want to be very specific when prompting Veo for video. From Google:

“Consider these elements when crafting your prompt:

• Subject and action: Clearly identify your characters or objects and describe their movements.
• Composition and camera motion: Frame your shot with terms like “wide shot” or “close-up,” and direct the camera with instructions like “tracking shot” or “aerial view.”
• Location and lighting: Don’t just name a place; paint a picture. The lighting and environment set the entire mood. Instead of “a room,” try describing “a dusty attic filled with forgotten treasures, a single beam of afternoon light cutting through a grimy window.”
• Alternative styles: Flow is not limited to realistic visual styles. You can explore a wide array of animation styles to match your story’s tone. Experiment with prompts that specify aesthetics like “stop motion,” “knitted animation” or “clay animation.”
• Audio and dialogue: While still an experimental feature, you can generate audio with your video by selecting Veo 3 in the model picker. You can then prompt the model to create ambient noise, specific sound effects, or even generate dialogue by including it in your prompt, optionally specifying details like tone, emotion, or accents. Note that speech is less likely to be generated if the requested dialogue doesn’t fit in the 8-second clip, or if it involves minors.

You can use Gemini to refine prompts, expand on an idea or be a brainstorming companion. Here’s a Gemini prompt to get you started:

“You are the world’s most intuitive visual communicator and expert prompt engineer. You possess a deep understanding of cinematic language, narrative structure, emotional resonance, the critical concept of filmic coverage and the specific capabilities of Google’s Veo AI model. Your mission is to transform my conceptual ideas into meticulously crafted, narrative-style text-to-video prompts that are visually breathtaking and technically precise for Veo.”

If you’re using Gemini to help generate multiple clips that have scene consistency, you’ll need to explicitly tell Gemini to repeat all essential details from prior prompts.”

My take: cheeky, prompting us to use Gemini to create prompts for Veo. Bit of a house of mirrors, no?