Start as a VTuber
from just one image

For an app released recently, the operating system may show a warning because it has little distribution history. There is nothing wrong with the app itself, but if you have any doubts, please do not force it open — contact us instead.

On Windows

1. When "Windows protected your PC" appears, click More info
2. Click the Run anyway button that appears

On macOS

1. Right-click the app icon (or control-click it)
2. Choose Open from the menu
3. Click Open once more in the confirmation dialog

No. Analysing your voice, and switching expressions based on what you say, are both done by AI running on your own PC. Neither the audio nor the text derived from it is sent to a server.

The AI data downloaded on first launch is what makes this on-device processing possible.

Yes. Generating 3D models, expressions, backgrounds and videos happens on external generative-AI services, not on your PC, so the image you upload (and intermediate images created during generation) is sent to the following services, only to the extent needed to run the generation.

• Image generation (expressions, backgrounds): OpenAI (United States)
• 3D model generation (Full body ver1 and ver3): Tripo (Hong Kong). For Full body ver3, we do not send the original illustration itself; we send only a character image redrawn for modeling by our image generation AI (OpenAI)
• 3D model generation (Full body ver2 and Bust), texturing for Full body ver3, rigging, motion, and texture repainting: Meshy (United States). Since 4 September 2026, however, we do not send the original illustration itself to Meshy; we send only character images redrawn for modeling by our image generation AI (OpenAI) and 3D model data
• Adjusting the head–body connection (only when automatic connection fails): Anthropic (USA). Only rendered images of the 3D model and shape measurements are sent, not the original illustration
• Video generation: BytePlus (Singapore)
• Checking whether an image is already published online (when an image is selected and before 3D generation starts): Google Cloud Vision (USA), TinEye (Canada) and SauceNAO (location not disclosed; images are deleted within 30 minutes)

Images are sent solely to run the generation, and are not passed to any service other than the above. The providers and their locations are also listed in the Privacy Policy.

We never use your images to train AI models. This service does not train anything — it calls external APIs in order to generate.

Here is what each service we send data to has published about training use (as of September 2026).

• OpenAI — API inputs and outputs are not used for training by default
• Tripo — its terms state that inputs and outputs are not used for training
• BytePlus — states that it does not use customer data for its own model training without the customer's prior authorisation
• Meshy — Its official API FAQ states that API inputs and generated models are not used for training.
  Note: On 7 September 2026, we reconfirmed this with a Meshy representative, who explained that API inputs and outputs are not used for training, regardless of whether we have an Enterprise agreement. We have therefore corrected our previous guidance that API data from non-Enterprise customers might be used for training. This service uses Meshy through its API.
• Anthropic — Its Commercial Terms state that it does not train models on customer content

These terms can change, so we review them regularly.

No. You may only use illustrations you drew yourself, or ones the rights holder has given you permission to use (Article 4 of the Terms of Service).

The copyright in the original illustration also extends to the 3D model generated from it. If you use or sell a model made from an image you have no rights to, the rights holder can demand that you stop, and claim damages.

Published images are checked automatically. To prevent unauthorised use of other people's illustrations, we check the input image against image-matching services (Google Cloud Vision, TinEye and SauceNAO) twice: when you select the image, and again before 3D generation starts. If the image is judged to be already published on the internet, generation does not start and no points are consumed.

An image counts as "already published" when the same image exists on the internet. Copies that have been resized or recompressed, cropped, or composited into a social-media post or an article banner are treated as the same image. A different drawing that merely has a similar style or composition, or a new drawing of the same character, is not affected.

Because the check cannot tell who the author is, this applies even to your own work: a published image cannot be used. If you want to generate from it, please prepare a new, unpublished image or ask the rights holder for a new drawing. The same goes for existing characters: please use a new, unpublished drawing commissioned from the rights holder.

Because the check compares against images each service has already collected, works that have not been collected may not be detected. The check is a supplementary safeguard; using images you do not hold the rights to is prohibited by the Terms of Service. If a rights holder files a claim, we will review it and act, including halting the generation and restricting the account. See the Privacy Policy for how images sent for matching are handled.

Please use the contact form at the bottom of this page and choose "Intellectual property claim". We accept claims from the rights holder or an authorised representative.

After reviewing the claim we will act on it, including halting the generations concerned and restricting the account involved.

Yes. Characters and images you generate can be used commercially or non-commercially.

This assumes you hold the rights you need to the source image. Please do not use illustrations drawn by someone else, or photographs of a person's face without their consent. See Articles 4 and 5 of the Terms of Service for details.

Yes. You may sell or distribute the 3D models you create. On the home screen, use "Saved characters" → "Export" to write the model, its animations and its expression textures into a single folder.

Crediting us is not required. If you would like to credit us anyway, we would be glad if you wrote "VTuber Original Character Maker".

If you sell a model, please pass these three points on to the buyer.

• Because of how generative AI works, a similar or identical model may be produced for another user. The design is not exclusive.
• We do not warrant that the model does not infringe the rights of third parties.
• Output from generative AI is not necessarily protected by copyright. Whether it is depends on the degree of creative involvement in making it.

See Article 5 of the Terms of Service for details. The exported folder also contains a text file saying the same thing.

No. Generation is driven by the illustration you upload, so another user generating from a different illustration will not, in practice, receive a model identical to yours — and even the same image gives a different result on every run. The models and images you create are never handed to other users, and are never reused as material for anyone else's generation.

That sentence, in Article 5 of the Terms of Service, is a legal disclaimer meaning we cannot guarantee that a design is exclusive. Output from generative AI can happen to resemble other output, and it is not always protected by copyright, so we cannot promise that nothing similar will ever exist anywhere. The notes we ask you to pass on to buyers when selling a model make the same point.

Feeding your illustration into this service without your permission is prohibited by Article 4 of the Terms of Service. If you find such a case, please use the contact form at the bottom of this page and choose "Intellectual property claim".

Points cannot be returned because of how a generation turned out. AI generation gives different results even from the same input, so a particular outcome cannot be guaranteed.

However, if a generation fails because of a system fault, the points spent are refunded automatically.

They expire 180 days after they are granted. Each purchase gets its own expiry date. Buying more later does not extend the life of points you already hold.

If you hold several, the ones expiring soonest are used first.

Open Options → OBS output at the top right and switch it on. The app then serves a URL that shows only the stream view. In OBS add Sources → + → Browser, paste that URL, and set the width to 1920 and the height to 1080. No plugin is needed.

None of the app's buttons or status panels appear in this feed. It carries video only, so add your microphone as a separate source in OBS. If you are using the default green-screen background, add Filters → Chroma Key in OBS to key out the green.

A window capture also works. See the guide for the steps.

A full-body image with the arms held away from the body works best. Images where the arms overlap the outfit, or where a sleeve and a skirt appear to merge, tend to come out misshapen when converted to 3D.

The background is removed automatically, so you can use the image as it is.

Downloading the app is free, and so is streaming with the character that comes bundled with it. That character already has expressions, the six mouth shapes used for lip sync, and an idle motion, so without generating anything you can try microphone-driven lip sync, topic-reactive expression switching, and the OBS output.

Creating your own original character requires buying points. 3D character generation, expression generation, background generation and video generation are the paid parts — see "Pricing" above for what each one costs.

To use the generation features you sign in with a Google account. This is how your point balance and the characters you create are tied to your account.