If Doubao Seedance 1.5 Pro had been released earlier, it would have saved me several hours. A few days ago, when I was working on AI video, the first frame and script didn't take much time, but creating audio-visual synchronized videos based on the first frame and script took a long time. Google Veo 3.1 has poor Chinese support, Sora 2 often mixes up the characters and their lines, Sun Wukong says Tang Sanzang's lines, and you need to draw cards multiple times before reaching the limit. Today I tested Doubao Seedance 1.5 Pro with the same images and prompts. Most shots were completed in one take, and the results were quite good. The characters and lines matched very well, and the lip movements were accurate. For example, my child really likes the panda in Figure 1 that speaks a dialect. The prompt is also very simple: just send the panda cub picture and the prompt below, and it will generate the image. The little panda ate bamboo shoots while saying in a sweet, childlike voice in Sichuan dialect, "Oh my, the sun is so warm and cozy!" I'll nibble on a bamboo shoot and slowly grow fat~.
This is also an example of a test I like to do: "What era are we living in? Still riding traditional brooms?" This is the best result I've achieved so far, a perfect one-shot. Prompt words: The video was generated based on a reference image. The shot uses a medium-shot tracking shot, focusing on the blonde witch in the foreground, showing her effortlessly and smoothly riding a white robotic vacuum cleaner through a bright sea of clouds, her cloak and hair billowing in the wind. The two witches in the background, riding traditional wooden brooms, appear unsteady and struggling, gradually falling behind. Around the 3-second mark, the blonde witch slightly turns her head to look back, a smug and mischievous smile on her face. Her lips move in sync with the dialogue: "What era are we living in? Still riding traditional brooms?" After saying this, she turns back to look ahead, and the robotic vacuum cleaner suddenly accelerates, leaving a brighter blue magical trail behind, leaving the witches in the background far behind, looking annoyed. The overall atmosphere is humorous and bright. Note that my prompts mentioned not only the dialogue but also camera movement, which is very useful. It allows for a variety of camera movements within a scene, such as long-take tracking shots and Hitchcock zooms, which significantly improves video quality.
This adorable little dragon video was made at my child's request. It was completed in three parts, each 5 seconds long, and then stitched together. The finished product is quite good, and my child likes it very much. The prompts for the three actions are as follows: Action Description: 1. A hand gently brushes the top of Xiaolong's head with a toothbrush. Xiaolong closes his eyes comfortably, looking very relaxed. 2. The little dragon shook its head, looked at the camera, and burped, unexpectedly spitting out a small ball of cute orange-red flame and smoke. 3. The little dragon grinned happily and tried to flap its little wings to fly up a little before comically falling back into the palm of my hand.
With the lower barrier to entry for video creation, making AI videos can be a lot of fun. I suggest you try it out on the Doubao App. Open the Doubao mobile app chat window, select "Animate Photos," select "Model 1.5pro" on the right, upload your image, and enter a prompt. You can then start making your video, which should take about 1-3 minutes.
