digitalhuman is a digital human video service for creating character speaking videos. Starting from facial images or videos, it produces lip-synced speaking videos. The creation interface provides inputs such as audio, text, and voice identifiers, with results delivered as video URLs and task information. It is suited to production workflows with existing character visuals and narration content, focusing on character expression rather than generating arbitrary scenes from scratch.
Clarify capacity, inputs and outputs, and invocation methods before selecting a model.
The above are digitalhuman's creation capabilities and invocation fields. Final video dimensions and duration are subject to actual results.
Learn what digitalhuman can bring to your work.
Use facial images or character videos as the starting point for creation, producing lip-synced videos around narration content. Ideal for projects with established on-camera character visuals that do not want to reshoot every time. Its core focus is character speaking; scripts, background design, and post-production packaging should still be arranged separately in the production workflow.
The interface provides an audio URL as well as text and voice identifier fields, making it easy to organize tasks around existing recordings or narration scripts. When using it, first determine which narration material to use, and avoid treating filling in all fields at once as a general practice; validate specific input combinations with samples before using them at scale.
Generation results include task identifiers, status, and video URLs, making them suitable for integration into content production workflows. Asynchronous options and callback addresses can be used to arrange result delivery; public MCP tools also provide task polling, batch retrieval, and deletion capabilities, allowing generation and subsequent organization to be handled separately.
Start with specific tasks to find where the model can be effective.
Prepare instructor visual assets and segmented scripts or recordings to create character narration clips for each knowledge point. The delivered videos can continue to be used for course editing, adding subtitles, and inserting presentation visuals. First verify the pronunciation and lip movements for technical terms, then process the full set of content to make it easier to maintain consistency in the final video.
Organize product feature descriptions into narration content and create introductory videos with a consistent presenter image. The resulting character narration clips can be combined with product screenshots, operation demonstrations, and brand assets. It is responsible for digital human presentation; interface operations, subtitles, or complete promotional video packaging should not be considered automatically included capabilities.
Continuously submit different narration content based on the same character assets to create program openings, announcements, or knowledge-sharing clips. After generation, save task identifiers and video URLs for archiving and later clip selection. Before scaling up in batches, first create a trial with representative content to check the voice, character performance, and editing compatibility.
Choose based on task complexity, input materials, and expected results.
If the task is to have a specified character explain content, digitalhuman's image or video input and lip-sync approach are better suited to the need. If the goal involves complex camera movement, changing environments, or multi-character narratives, prioritize scene-generation video capabilities; do not treat character narration and general video generation as the same creative approach.
When recordings are already available, tasks can be organized around the audio URL; when starting with a script, you can try a combination of text and voice identifiers. For custom voices, you can separately prepare them using MCP's voice cloning capability. Do not interpret the service as different performance tiers based on the old engine field; determine the production plan based on the actual character assets and narration results.
From a small-scale task to formal integration.
Define the objective, required inputs, and output requirements, using real business examples as a starting point.
Open the trial page, confirm the parameters supported by this entry point, then submit a small-scale task to review the results.
Keep the complete model ID, use the request format specified in the documentation, and confirm billing rules on the Pricing page.
Before formal use, understand the output quality and capability scope.
Answers to common questions when using digitalhuman.
It is mainly used for lip-synced narration based on face images or videos. The text field is used for narration content and should not be understood as being able to generate arbitrary videos solely from scene descriptions. To create complex environments, shots, or stories, choose a more suitable creation method.
They do not have to be limited to videos; digital human creation also supports starting from face images. The API provides image_url and video_url respectively, which can be prepared according to the asset type. For first-time use, it is recommended to first create a short test clip, confirm the character's performance, and then use the same assets for subsequent tasks.
The API provides audio_url, as well as text and voice_id fields, allowing creation around recordings or scripts. The two approaches should not be mixed without validation; first choose the narration method, test the corresponding asset combination, and then use successful configurations for ongoing production.
The supporting MCP tool supports cloning voices through short reference audio, but the video generation API does not have a dedicated field for cloning samples. Voice preparation should be completed separately before arranging narration generation; do not confuse character-driving audio with voice cloning reference materials.
Submit a task through POST /digital-human/videos; the response provides the video URL and task-related fields. When using an asynchronous workflow, you can use callbacks or MCP task queries to obtain progress, determine from the returned status whether a usable result is available, and then download the video for post-production.
Model information · Updated: 2026-10-01. For invocation parameters and billing rules, see the API and pricing sections.
Start with a clear goal, and judge whether it fits your work based on real results.
This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs