Animate Text

Animate an image using audio. The request is a multi-part form data POST that takes the following arguments

• init_image The image you want to animate

• driving_text The text that the animated character should talk

• audio_seed The audio seed that should be used to find the voice. Different seeds have different voices

• user_id: A GUID that identifies the caller of the API. Please ensure the same user_id is passed in all requests. If you have multiple products using the API, you can use different user_id for different products. But please do not generate a new user_id in every request.

• request_id: A GUID identifying your request

• seamless_loop: Loop the result video so that it can loop seamlessly. This is useful for NPC characters in Games. Also looks nice.

• should_add_audio: Should audio from the driving video be added back to result?

• remove_bg: (Experimental) Try to remove the background in result video

• generated_animated_frames_as_png: Generate the frames of the result video as PNG frames. These can be huge and be in the order of several GBs.

• upscale_video_4x: Upscale the result video by 4x. So if input is 512x512px image, this will result in 2048x2048 video.

• make_slowmo_animation: Make the result in slow motion.

• smoother_animation_2x: Double the number of frames in the result video. Gives smoother animation.

• crop_animated_video_to_face: For large input images with full body or half body of the person, should the result video be cropped to the persons face?

• dummy_image: A dummy image is used if the facial features of the input image init_image are not clear. Pass in another image that has clear facial features in the same position as the init_image

• expressiveness: The amount of mouth movement in the animation, 1.0 has minimal movement, 2.5 has lots.

• still_image: If true, the image will not turn around or move while speaking

• animation_mode: Animation mode defines how the result animated video should look like. This can take three options:

  standard - Make the result video look like the driving video

  relative - Make the result video look like the input image

  make_init_frame - Choose the best possible mix of the two modes above