Tips & Tricks
Best practices and creative ideas for creating high quality videos with Roast AI.
Recommendation 1: Use single-speaker videos
Select videos with a single speaker whose face is clearly visible. This will result in higher quality lip-movements and facial expressions. For example, a selfie-video of yourself talking directly to the camera will work well. We use state of the art AI models to realistically synch the audio and video.
As the underlying models improve, they will allow for more complex, multi-speaker scenarios. Currently, we suggest using videos with only a single speaker to ensure the best results and reduce artifacts.
Recommendation 2: Use videos with at least one minute of clear audio
Clear audio is crutial to creating a realistic voice-clone. Around one-minute of video with minimal background noise produces the best results for our models.
Avoid background noise like cars, background voices, music, or mechanican humming. The video you input is the sole source of the voice-clone, so make sure it has the tone and audio quality you want.
Recommendation 3: Add appropriate disclaimers
With great power comes great responsibility. Make sure to add the appropriate disclaimers to your videos depending on your type of content and how you intend to share it. You can find more guidance about acceptable use in our terms of service.
We provide a suite of features such as adding watermarks and disclaimers to make it easy for users to comply with deepfake regulations. When in doubt, read the guidelines for your jurisdiction, use your best judgement, and act responsibly.