
Alibaba Cloud's platform for making...
For a training voiceover, choose landscape or portrait first, because the ratio cannot be changed after the project is created. Look and voice can come from the public library or from your own image and voice, and text or audio decides whether the digital human reads a script or follows a recording. The background can be a built-in image, a solid color, or an upload, and the figure can be dragged and scaled in the preview. Progress and download sit in the generation history, and the project can be edited again before another export. The free trial covers only a limited duration, and full use requires the service to be activated. Realtime conversation can be called from a device toolkit only after concurrency is purchased. It fits teams that want a digital human instead of a person on camera for courses, livestreams, and tutoring.
Tongyi Lingmou is Alibaba Cloud's platform for making and using digital humans. The help center now describes the same product as Tongyi Wanxiang digital humans. Users can train a look from video, an image, or a 3D asset, or pick a public 2D or 3D look, and clone a voice. Video creation takes typed text or an uploaded or recorded audio file, a male, female, or custom voice, then a landscape or portrait frame and a background, and produces a downloadable presenter video. Realtime conversation combines speech recognition, speech synthesis, a chat model, and digital-human driving for training, livestream selling, and tutoring.
Difficulty: Intermediate
Custom look and voice
A look can be trained from video, an image, or a 3D asset, or taken from the public library, with a voice clone close to the original.
Speak from text or audio
Type the words and preview them, or upload or record audio, and the chosen voice drives the look into a video.
Frame and background
Choose landscape or portrait at creation, after which the ratio stays fixed; the background can be built-in, a solid color, or an upload.
Realtime conversation project
Speech recognition, speech synthesis, a chat model, and digital-human driving sit in one conversation project that can be tried in the console.
Course presenter video
A trainer picks a portrait or landscape look, types the script, previews the voice, and exports a downloadable course clip.
Read a script in your own voice
Clone a voice first, then have a public or custom look read a new product explanation in that voice.
Try a conversation in the console
Talk with an official conversation look on the realtime page, check the mouth and the reply, then connect it to your own app.
Yes. The docs say an image can be trained immediately into a presenter video, without a studio shoot. A public 2D or 3D look can be selected instead, then paired with a male, female, or cloned voice. If a recording already exists, uploaded audio can drive the video.
The video guide says the free trial only supports clips of limited length, and full use needs the service activated. Finished videos are viewed and downloaded from the generation history, and a project can be edited again before another export. Connecting realtime conversation to your own app also requires purchased concurrency. A console chat and a device call are not the same step.
Real reviews and feedback from users