# Digital human: the complete guide

> Updated: 2026-10-29 · Applies to KiiKii
> HTML version: https://kiikii.cc/en/doc/digital-human

## Quick answer

- Five modes: Standard, One-Take, Music MV, E-commerce Selling, and Store Visit.
- Standard runs 30–150 s, One-Take 20–60 s, Music MV follows the song, and E-commerce Selling and Store Visit run 20–80 s.
- Cloned avatars need 1–3 clear frontal portraits; designed avatars create a new IP from text and presets.
- Voice cloning needs 10–60 seconds of clear speech in wav / mp3 / m4a under 10 MB.
- Full flow: choose avatar and voice → describe the brief → confirm the shot plan → generate → preview and download.

## 1. Understanding digital humans
A digital human links avatar, voice, script, and shot planning into one production line: you choose an avatar and a voice, describe what the video should say, and the AI plans the shots, generates the voiceover and visuals, then stitches a downloadable final video. Output types include voiceover explainers, multi-shot videos, music MVs, e-commerce selling clips, store-visit promotions, and a personal digital twin. Creation is two-stage—plan first, generate second—and planning does not consume video generation time.

## 2. Which industries and users it suits
It fits wherever you need a consistent on-camera presence, high-volume output, and low-cost updates: knowledge and education, e-commerce retail and brand marketing, local life and store visits, healthcare and financial services, corporate and government communications, creators and MCNs. Typical users are founders who prefer not to appear, instructors and knowledge creators, e-commerce and livestream teams, store owners, marketing teams, and cross-border multilingual teams.

## 3. Five creation modes
All modes support 9:16 and 16:9 (default 9:16) and 480p / 720p / 1080p (default 720p); except MV, each lets you set a target duration or leave it to the AI. **Standard** (30–150 s, up to 24 shots) plans hook, point, demo, evidence, selling point, and call-to-action shots in mixed formats. **One-Take** (20–60 s) stitches 5–15 s segments so the person, scene, and tone stay consistent. **Music MV** follows the song for up to 60 shots, with an optional second avatar. **E-commerce Selling** (20–80 s) builds from product and scene images around pain point, benefit, offer, and call to action. **Store Visit** (20–80 s) uses up to 3 scene images following storefront → interior → signature → price or address.

## 4. Supported durations and generation settings
Auto duration is decided by the AI from information volume: about 20–55 s for Standard and 8–28 s for One-Take. Final videos may deviate slightly because of per-shot trimming, voiceover length, and stitching, which is normal. Frame 9:16 suits Douyin, WeChat Channels, and Xiaohongshu; 16:9 suits websites, courses, and launches. Resolution runs 480p for previews, 720p everyday, and 1080p for large screens; burn-in subtitles write subtitles into the picture, while MV lyrics always follow the timeline and are never burned in.

## 5. Cloned avatars and designed avatars
A **cloned avatar** takes 1–3 clear frontal portraits; the AI preserves facial features, face shape, skin tone, and character, then generates a 16:9 four-view reference sheet (front / side / back / facial close-up), with gender, age, height, and weight required and cloning usually taking 1–3 minutes. A **designed avatar** needs no photo: choose gender, nationality style, face, expression, hairstyle, outfit, pose, age, height, and weight to build a new IP, optionally with outfit or hairstyle references. Both save to My Avatars; avatars can be submitted for public listing with up to 3 category tags, and failed review leaves them fully usable as private avatars.

## 6. System voices and voice cloning
Hundreds of system voices can be filtered by gender, age, language, and scenario across Chinese, English, and mixed Chinese–English; they need no sample and suit batch testing and multi-character content. Voice cloning takes a 10–60 s clear sample (wav / mp3 / m4a, ≤10 MB, quiet room, single speaker, natural pacing, no background music), validates it, binds the Qwen synthesis model, and adds the result to My Voices. Each user can hold up to 10 cloned voices, usable in digital human voiceover, short drama dubbing, and music, and an A/B comparison plays the original sample against the cloned voice.

## 7. From brief to final video
A good brief covers five things: topic and audience, structure and key points, duration and frame, visuals and assets (referenced as @Image 1), and a closing call to action; the page provides a reusable 60-second template. After submission the AI enters Planning, proposes a shot script and lines, then waits at Pending Confirmation, where you can review lines, visual descriptions, camera moves, durations, and subtitle style, revise the copy conversationally, or regenerate a single shot. Generation runs in the background and survives leaving the page; if one segment fails, retry only that segment rather than the whole video.

## 8. Demo videos
Two final videos from real creation records: "Understanding Chinese History in Thirty Seconds" (Standard mode, 9:16, 720p, 4 shots) and "Calamine Soap Special Offer ¥1.9" (E-commerce Selling, 26 seconds, 3 shots, high-energy bouncing subtitles).

## 9. Get started
For a first try use Standard mode + Auto duration + 9:16 + 720p, and aim to ship a 30-second voiceover rather than complex shot planning. Pick a public avatar or clone one from 1–3 frontal photos, preview system voices (clone only when you need recognition), write a 60-second brief with the five-part template, confirm the plan and generate, then review the first 3 seconds, subtitle readability, and the closing call to action.

## FAQ

### Why doesn't the final video length exactly match the setting?
Videos are assembled from 5–15 second shots, and voiceover length, per-shot trimming, and local animation affect duration; in Auto mode the AI plans by information volume, so small deviations are expected. MV duration is set by the song.
### Is One-Take really generated in a single pass?
It looks continuous, but under the hood it generates 5–15 second segments and stitches them, keeping person, scene, and tone consistent; above 15 seconds it splits into 2–3 continuous generations.
### What if the digital human doesn't look enough like me?
Use 1–3 frontal portraits that are clear, unobstructed, and evenly lit; avoid sunglasses, profile angles, heavy filters, and group photos. A cloned avatar follows the reference portraits, so a clearer photo beats more text.
### Why don't the clothes in the video match the avatar image?
Check whether Clothing Consistency is enabled. When on, the digital human wears exactly what the avatar image shows; when off, the AI designs outfits per scene while keeping the face consistent.
### How long a sample does voice cloning need?
10–60 seconds of clear speech in wav / mp3 / m4a, no larger than 10 MB, recorded in a quiet room by a single speaker at natural pacing without background music; an A/B comparison is available after cloning.
### Can I request to publish an avatar? Can I still use it if review fails?
Yes. Add up to 3 category tags and submit for review; approved avatars enter the public library. Avatars that fail review show the reason but remain fully usable as private avatars.
### What if the video has no audio or the lip sync is off?
Confirm that a voice is selected and previews normally, then check whether that shot generated successfully. Retry a failed segment in the shot details; if the whole result is broken, regenerate the final video.
### Can the generated videos be used commercially?
Make sure you hold the appropriate rights to the avatar, voice, images, music, and script, and comply with platform rules and the law. Do not clone others' avatars or voices without authorization or use generated content to mislead.