Duix
duix.com
What it is
Duix is a conversational AI platform for building and using interactive avatars with faces and voices. Users can customize avatar appearance, voice, language, persona and conversation logic, and connect knowledge bases. Its visual capabilities include lip-sync, facial rendering and body motion, as well as recognition of faces, gestures, emotions and environmental context. Listed use cases range from customer service, interviews and school AI to AI doctor and personal conversational experiences. The homepage claims 200ms TTFB latency for real-time two-way interactions. Duix supports Android, iOS, web and API access, and says every plan includes API access; starting or using an API session requires conversation minutes in the account. The Free plan is free and includes 5 minutes of talk time per day, one concurrent stream and stock avatars, but no custom AI avatar support. Paid plans start at $29/mo. The pricing page says annual packages are not currently available.
Spec sheet
- What it does
- Duix presents itself as conversational AI with realistic faces and voices for interactive digital-human applications.Read 14 Oct 2026 · duix.com
- Product
- Duix describes itself as a conversational AI platform with real faces and voices for interactive AI avatars.Read 7 Oct 2026 · duix.com
- Support
- Duix lists support, business, and partnership inquiries at [email protected].Read 7 Oct 2026 · duix.com
- Maker
- Duix describes itself as an AI and human-computer interaction research lab with over 100 patents and says it has an international team in Singapore.Read 7 Oct 2026 · duix.com
- API access
- Duix documentation states that every plan, including Free, provides API access.Read 14 Oct 2026 · docs.duix.com
- API billing
- API sessions require available conversation minutes, supplied by a paid plan or the free quota.Read 14 Oct 2026 · docs.duix.com
- API usage limit
- Using the API to start or use a session requires available conversation minutes in the account.Read 7 Oct 2026 · docs.duix.com
- Avatar realism
- The product page describes lip-sync, facial rendering and body-motion rendering as core realism capabilities.Read 14 Oct 2026 · duix.com
- Backers
- The About page says Duix is backed by Sequoia Capital and Tencent Capital.Read 7 Oct 2026 · duix.com
- Commercial cloud licensing
- The pricing page distinguishes personal licenses on lower plans from commercial licenses on Scale, Growth, Business and Enterprise.Read 14 Oct 2026 · duix.com
- Company founding
- LinkedIn lists the company as founded in 2017.Read 14 Oct 2026 · hk.linkedin.com
- Company headquarters
- LinkedIn lists SMART INTELLIGENCE PTE. LTD.'s headquarters in Singapore.Read 14 Oct 2026 · hk.linkedin.com
- Company size
- LinkedIn lists the company size as 501–1,000 employees.Read 14 Oct 2026 · hk.linkedin.com
- Customization
- The homepage says developers can customize avatars, voices, personas and conversation logic through APIs and SDKs.Read 14 Oct 2026 · duix.com
- Developer portal
- The developer portal manages API keys, conversations and avatars and lets users test configurations before integration.Read 14 Oct 2026 · docs.duix.com
- Documentation depth
- The public docs include getting-started guidance, API reference material, conversation-ID instructions and links to community channels.Read 14 Oct 2026 · docs.duix.com
- Funding stated by maker
- The About page says Duix is backed by Sequoia Capital and Tencent Capital.Read 14 Oct 2026 · duix.com
- Independent platform deployment
- AI Directories reports deployment targets of Android, iOS and web, with APIs for sessions, conversations and interaction data.Read 14 Oct 2026 · aidirectori.es
- Independent platform description
- AI Directories describes DUIX as a cloud platform with standardized APIs and SDKs for vision, voice, emotion and conversation functions.Read 14 Oct 2026 · aidirectori.es
- Independent praise: output and speed
- A SegmentFault reviewer praised fast rendering, broad photo compatibility and natural front-facing lip-sync for batch talking-head videos.Read 14 Oct 2026 · segmentfault.com
- Independent praise: setup
- The SegmentFault review reported that installation was comparatively simple and that a Windows one-click package was available.Read 14 Oct 2026 · segmentfault.com
- Integration
- Duix says it supports SDKs and APIs, with customization for avatars, voices, personas, and conversation logic.Read 7 Oct 2026 · duix.com
- Knowledge integration
- Duix says its conversational avatars can integrate with knowledge bases.Read 14 Oct 2026 · duix.com
- Legal restrictions
- The terms prohibit reverse engineering, sublicensing, unauthorized access and using the platform to build a competing service.Read 14 Oct 2026 · duix.com
- Limit: body movement
- The independent review says Duix-Avatar drives the head and shoulders rather than full-body gestures or movement.Read 14 Oct 2026 · segmentfault.com
- Limit: environment
- The review notes that Python-version matching, FFmpeg and CUDA/cuDNN dependencies can still challenge beginners, especially on Mac and Linux.Read 14 Oct 2026 · segmentfault.com
- Limit: head motion
- The SegmentFault reviewer found head movement limited to small nods and subtle motion, making longer clips feel stiff.Read 14 Oct 2026 · segmentfault.com
- Limit: profile faces
- The same review reported noticeably worse results with 30-, 45- and 60-degree profile images than with front-facing photos.Read 14 Oct 2026 · segmentfault.com
- Linux requirements
- The documented Ubuntu deployment recommends Ubuntu 22.04, at least 32 GB RAM, an RTX 4070 and more than 100 GB of disk space.Read 14 Oct 2026 · github.com
- Local APIs
- The repository documents local model-training, audio-synthesis and video-synthesis HTTP endpoints exposed by Docker services.Read 14 Oct 2026 · github.com
- Maker company name
- LinkedIn identifies the maker as SMART INTELLIGENCE PTE. LTD.Read 14 Oct 2026 · hk.linkedin.com
- Multilingual support
- Duix says its platform supports more than 50 languages on its paid plans.Read 14 Oct 2026 · duix.com
- Notable company disclosure
- A 2025 Hong Kong Exchange filing says the broader GJ AI business launched its Yan Di large language model in 2023 and developed the DUIX ONE multimodal engine.Read 14 Oct 2026 · www1.hkexnews.hk
- Notable limit
- The pricing page says annual packages are not available at the moment.Read 7 Oct 2026 · duix.com
- Official research background
- Duix describes itself as an AI and human-computer-interaction research lab with more than 100 patents.Read 14 Oct 2026 · duix.com
- Open-source license
- An independent SegmentFault review identifies the Duix-Avatar repository license as Apache-2.0.Read 14 Oct 2026 · segmentfault.com
- Open-source operational limit
- The repository says local deployment requires an NVIDIA GPU and drivers; services will not start without them.Read 14 Oct 2026 · github.com
- Open-source project
- The official GitHub repository describes Duix.Avatar as a free, open-source toolkit for offline video generation and digital-human cloning.Read 14 Oct 2026 · github.com
- Privacy and data
- Duix's terms say it may use customer content, media files, and interactive media files to provide, maintain, and improve its products and services.Read 7 Oct 2026 · duix.com
- Rating on SegmentFault
- The independent review gives Duix-Avatar an overall score of 4.2 out of 5.0.Read 14 Oct 2026 · segmentfault.com
- Real time
- The homepage claims 200ms TTFB latency for real-time two-way interactions.Read 7 Oct 2026 · duix.com
- Real-time latency
- The homepage claims 200 ms time to first byte for two-way avatar interactions.Read 14 Oct 2026 · duix.com
- SDK relationship
- The documentation says the web application uses the same APIs as the SDK, although generation is nondeterministic.Read 14 Oct 2026 · docs.duix.com
- Self-hosting
- The repository documents local Docker deployment for Windows and Ubuntu 22.04 and says processing is offline.Read 14 Oct 2026 · github.com
- Singapore team
- Duix says it established an international team in Singapore for localized support and compliance assurance.Read 14 Oct 2026 · duix.com
- Use cases
- The official site lists interactive avatars, AI doctors, customer service, interviews, AI companions and school assistants.Read 14 Oct 2026 · duix.com
- Vision
- Duix says its vision functions can recognize faces, gestures, emotions and environmental context.Read 14 Oct 2026 · duix.com
- Visual features
- Duix describes lip-sync, facial rendering, body motion, and vision capabilities for recognizing faces, gestures, emotions, and environmental context.Read 7 Oct 2026 · duix.com
- Windows requirements
- The repository requires Windows 10 build 19042.1526 or later, more than 30 GB on D:, more than 100 GB on C:, and recommends an RTX 4070 with 32 GB RAM.Read 14 Oct 2026 · github.com
Plans and prices
Strong points and limits
Strong points
- Free plan includes 5 minutes of talk time daily
- API access is available on every plan
- Supports avatar customization and knowledge-base connections
- Lists 50+ languages and 1080p output
Limits
- Free plan does not support custom AI avatars
- API sessions require available conversation minutes
- Annual packages are not currently available
Who it suits
Duix suits teams or individuals seeking interactive AI avatars for the listed conversational use cases. API access is available on every plan, but API sessions require conversation minutes.
Duix in depth
Read the full write-up
Duix is a platform for building live conversations with AI avatars, suited to developers and service teams that want avatars inside an application. Its strongest case is the combination of API access and adjustable personas; its main constraint is that every session draws from a capped monthly or daily talk-time allowance.
Overview
Duix brings faces, voices and conversational logic together for interactive avatar experiences. Its use cases span customer service, interviews, school and AI doctor experiences, as well as personal companion applications. It claims 200ms TTFB latency for real-time two-way interactions and describes lip-sync, facial rendering, body motion and vision features that recognize faces, gestures, emotions and environmental context. These capabilities suit responsive conversations better than a workflow focused solely on producing finished avatar videos.
Users can customize avatar appearance and voice, support multiple languages and connect knowledge bases. SDKs and APIs enable integration into applications, with room to tailor avatars, voices, personas and conversation logic. Duix says it may use customer content, media files and interactive media files to provide, maintain and improve its products and services, a consideration for teams working with sensitive material.
Key features
Interactive avatars: Two-way conversations, facial rendering, lip-sync and body motion are aimed at experiences that respond to a user, not just pre-rendered clips. The homepage's 200ms TTFB claim is appealing for conversational use, though it is a stated figure rather than a deployment guarantee.
Customization and languages: Appearance, voice, persona and conversation logic can be tailored; the plans include support for 50+ languages. Connecting a knowledge base can ground an avatar in an organization's material, while stock avatars give buyers a starting point without creating one.
API and SDK access: API access appears on every plan, including Free, making it possible to evaluate application integration before paying. It is not a separate pool of usage: starting or using an API session requires available conversation minutes.
Vision capabilities: Duix describes recognition of faces, gestures, emotions and environmental context. That broadens the intended interaction beyond voice alone, but the value depends on whether an application needs those visual cues.
Pricing
Plans are billed monthly, and annual packages are not currently available. Free offers 5 minutes of talk time per day, one concurrent stream, stock avatars, API access, skill integration, 50+ languages and 1080p output. It carries a personal license and does not support custom AI avatars. This is a useful low-commitment way to explore the platform, but the daily allowance and single stream make it a poor fit for sustained or multi-user deployment.
Starter costs $29 a month and raises the allowance to 150 minutes per month. It keeps one concurrent stream and a personal license, and still excludes custom avatars. It suits a small personal project with modest monthly usage, but its minutes are a fixed monthly pool rather than a daily allowance.
Creator Popular costs $59 a month and includes 300 minutes, three personal AI avatar creations per month and one concurrent stream, alongside the same personal license, API access, skill integration, languages and 1080p output. It is the entry point for buyers who need custom avatars, though the single stream remains a bottleneck for simultaneous conversations.
Scale costs $399 a month for 3,000 minutes, seven avatar creations per month and five concurrent streams, with a commercial license. Growth costs $799 a month for 6,500 minutes, ten creations and 12 streams. Business costs $1,999 a month for 20,000 minutes, 20 creations and 20 streams. These tiers move from personal experimentation to commercial use, adding both substantial capacity and concurrency; buyers should match the higher monthly commitment to expected session volume.
Enterprise costs $4,999 a month for 60,000 minutes, 50 avatar creations, 50 concurrent streams and technical personnel support, as well as a commercial license. It is the clearest fit for high-volume services that need substantial simultaneous capacity. Customization has custom pricing, with talk time, concurrency and avatar creation arranged on demand, for requirements that do not fit the fixed tiers.
Platforms
Duix is available on web, Android and iOS, and offers an API for application integration. Every plan supports output up to 1080p; the listed plans do not offer higher resolution.
Who it's for
Duix is most compelling for developers and organizations adding conversational avatars to customer service, education, interview or other interactive applications. The free plan can support a brief evaluation, while Creator Popular is the first tier with custom avatar creation. Commercial deployments should look to Scale or above for a commercial license and more than one concurrent stream.
It is less suited to buyers who only need a small number of finished avatar videos, or who require generous concurrent use at a low monthly price. Teams handling sensitive media should also weigh Duix's stated use of customer and interactive content to improve its services.
Pros and cons
Pro: API access is included even on Free, so application integration can be explored without first purchasing a paid tier.
Pro: Plans combine 50+ languages, stock avatars, skill integration and 1080p output, while paid tiers add custom avatar creation and commercial licensing.
Pro: Scale and higher plans increase both minutes and concurrent streams, supporting larger live services.
Con: API sessions consume the same conversation-minute allowance, so integration does not bypass usage caps.
Con: Free and Starter exclude custom avatars, while Starter and Creator Popular allow only one concurrent stream.
Con: Commercial licensing starts at Scale, at $399 a month; higher concurrency and volume require materially more expensive plans.
Con: Annual packages are unavailable, and the terms permit use of customer media to provide, maintain and improve services.
Alternatives
For a broader comparison of products in this category, see AI Avatar Video Software and AI Avatar Generators.
HeyGen is another freemium option on web, Android, iOS and API; its free plan allows three videos per month, each up to one minute, making it a better match for limited finished-video output.
Makefun AI offers a free plan with daily credits, 720p video, watermarking and ten custom avatars, a different fit for buyers prioritizing avatar creation over Duix's live conversation model.
Vimotion offers a free first-video package with a hero video up to 60 seconds and two social cuts; consider it for a small, packaged video deliverable rather than ongoing conversational sessions.
AiVATAR is a web-based alternative with a one-time 600-credit free allotment and one minute of video, suited to a brief video-generation tryout.
DeepFA AI Avatar is another freemium avatar option for buyers comparing an alternative service.
Rendora offers a freemium plan with monthly credits, a three-minute video maximum and 540p output, which may suit short-form video generation rather than interactive sessions.
Cuevo AI has a free plan with daily points and 1080p video; it is worth considering when free video rendering is the priority.
LipSynthesis offers one minute of free video generation, a straightforward option for a limited video-generation need.
Verdict
Choose Duix if you need customizable, multilingual avatars for live conversations and want API access from the outset. Its strongest advantage is the path from free API evaluation to commercial plans with larger minute pools and concurrency. Look elsewhere if you need only finished avatar videos, require more than one stream on an entry plan, or cannot accept the monthly usage caps and media-use terms.
Verdict
Duix pairs customizable conversational avatars with API access and a free entry plan. Compare the included talk time and concurrency with the intended usage before choosing a paid tier.
Where Specifiction places it
- AI Avatar Video Software#1 of 279.3
- AI Avatar Generators#5 of 289.2