
- Monthly visits
- 223
- 2026-10
- Monthly growth +7,333.3%
daVinci-MagiHuman - Turn a portrait and text or audio into a lip-synced talking video
daVinci-MagiHuman is a 15-billion-parameter open-source AI model for generating talking-head videos. It takes a portrait photo and text or audio, then generates synchronized speech and video. The model is released under the Apache 2.0 license for local inference and commercial use subject to its terms. An online studio is also available; its subscriptions and credit options are priced separately.
- AI Video
- AI Avatars
- AI talking video generator
- portrait lip sync model
- open-source audio video AI




Capabilities
Talking-video generation
Generates a lip-synced talking video with matching audio. Input a face portrait and a text script or audio; output resolution and inference speed depend on the settings and available hardware.
Joint audio-video generation
Generates audio and video together in one model pass. The single-stream Transformer processes audio and video tokens jointly rather than requiring separate text-to-speech and video pipelines.
Local model inference
Runs inference locally using the released model weights. Download the checkpoints and follow the upstream README for setup, inference scripts, and command-line options; hardware affects throughput.
Online video studio
Provides a hosted workflow for generating AI videos. Users can access the studio online, with subscriptions and separately purchasable credits offered on the pricing page.
Pricing
The hosted studio offers free access, recurring monthly or yearly subscriptions, and one-time credit purchases. Compare credit amounts, daily free images, concurrent tasks, character limits, support, and model access when choosing.
| Plan | Price | Included features and limits |
|---|---|---|
| Basic | $10/month |
|
| Pro | $20/month |
|
| Max | $50/month |
|
| Enterprise | Flexible plans; contact sales |
|
| Pay-as-you-go credits | One-time purchase; price varies |
|
Yearly subscriptions are billed once for the full year at a discounted effective rate; the page advertises 20% savings. Subscriptions renew automatically. New subscriptions have a seven-day refund window; renewal charge concerns must be raised within 72 hours.
Use cases
Create a lip-synced talking-head clip
When a portrait needs to speak a script or supplied audio, provide the portrait and text or audio as inputs. The model generates a talking video with speech and lip motion aligned.
Run talking-video inference on your hardware
When a developer or researcher wants to run the model locally or on a server, obtain the checkpoints from Hugging Face and use the project README to configure inference. The resulting generation runs on the selected hardware.
Generate videos in the hosted studio
When a user wants an online generation workflow rather than installing the model, open the AI video studio and submit a supported generation job. The pricing page describes subscriptions, credits, and credit-based generation costs.
Limitations
The hosted subscription credit balance pauses requests when it reaches zero; users can add pay-as-you-go credits or upgrade to resume generation.
The model's supported lip-sync languages depend on training data and the released model; the current language list is referred to the official README.
Inference speed depends on settings and hardware; the approximately two-second figure applies to a reported two-second 256p clip on an H100-class GPU.
Commercial use of the model is subject to Apache 2.0 license conditions, including applicable notices and attribution requirements.
Pay-as-you-go credits do not expire, but subscription credits are provided on a monthly or yearly billing cycle.
Who it is for
Developers integrating or running an open talking-video model
Researchers evaluating audio-video generation models and inference
Creators who need to generate portrait-based talking videos online
FAQ
What does daVinci-MagiHuman generate?
It generates a lip-synced talking video with matching audio from a portrait and text or audio. The model jointly generates the audio and video.
Is the model free to use?
The released model weights and code are available under Apache 2.0. Hosted studio subscriptions and credits have separate pricing and terms.
Can I use daVinci-MagiHuman commercially?
Yes, subject to Apache 2.0 license conditions. Review the license requirements, including notices and attribution, for your use.
Which languages does it support for lip sync?
Language coverage depends on the released model and its training data. Check the official README for the current supported languages and locale-specific caveats.
How are online studio credits charged?
Generation uses credits based on output length and settings such as resolution and mode. The pricing FAQ gives approximate examples and advises checking live totals in the studio.
Do pay-as-you-go credits expire?
No. Credits purchased through the one-time option do not expire, unlike subscription credits that are issued on a monthly or yearly cycle.
Comments