I was just scrolling through my feed and BAM! My mind was blown by an incredible post from Tianyu Xu.
Ever get frustrated that your AI video characters are stuck speaking English? Well, Tianyu just showed everyone how to break that barrier with Google’s Veo 3. It’s simpler than you think!
Here’s the super-simple trick he discovered:
- 1️⃣ Tell Veo the Language: Start your prompt by explicitly stating which language the character will speak (e.g., “…and says in Mandarin:”).
- 2️⃣ Write the Script: In the dialogue part of your prompt, just type out the script in your target language.
- 3️⃣ Use the API: Tianyu notes you’ll get a much higher success rate using the Gemini/Veo API instead of the Flow interface.
He even dropped a perfect example prompt to show it in action:
Camera cuts to a close-up shot of the man’s face. He raises his eyebrow with a mysterious smile and says in Mandarin: 四是四, 十是十, 十四是十四, 四十是四十.
Known Issues to Watch Out For:
- 📌 Flow Rejects Prompts: The standard Flow interface might reject prompts with non-Latin characters.
- 📌 No Audio Glitches: Sometimes the audio just doesn’t generate, which can happen in both Flow and the API.
- 📌 Surprise Accents: The model might sometimes generate an unexpected accent.
He also used this test to push Veo 3’s facial expressions and was pretty happy with his avatar’s acting skills. This is a total game-changer for creating more diverse and global content!
Seriously, go check out Tianyu Xu’s original post to see the video he made with this technique. It’s awesome stuff!