Project (NDA)
AI avatar for streams
NDA: on air instead of the camera face - a chosen character or persona. AI changes the host’s appearance in realtime; the page has a demo of how it looks.
live face swap
Format
realtime stream
Mode
NDA / no brand
Scope
video on the page
Demo
- 01 · Client
Client - a B2C streaming project (NDA).
- 02 · Challenge
Face replacement live, not a pre-rendered video.
- 03 · Solution
We built an AI broadcast tool: on camera - a person, on air - the chosen character.
- 04 · Result
Implemented live face swap for streams: on air instead of the camera face - a chosen character or persona, change in realtime.
AI face swap for streams | JimmyNeuron case
About the client
Client - a B2C streaming project (NDA). They needed a live tool: the host turns on the camera, picks a character - and viewers see not their face, but the chosen look.
Not a Stories filter and not offline video render: appearance swap runs in the live stream together with character/persona select.
The challenge we started with
- Face replacement live, not a pre-rendered video
- Choice of character / persona instead of one hard-coded mask
- Output into the usual streaming pipeline without breaking the broadcast
- NDA: no public details of the brand or integrations
How it works on air
The host starts a stream from the camera and picks a character or persona. On the live broadcast, viewers see the chosen look instead of the face of the person on camera. That is the essence of the product - an AI broadcast with appearance swap in realtime.
Character selection
The look is not fixed forever: the character/persona can be changed to fit the show format. This includes cross-gender and stage scenarios - a man can host in a female look and vice versa.
Live, not offline
The pipeline is built for live broadcast: the swap runs together with the camera, with a target smoothness of ~60 fps. Not "generate a clip and publish it" - a streaming tool.
NDA and demo
The brand and commercial details are covered by an NDA. In this case we show only a demo of the mechanics: how the face swap and character selection look. There is no public link to the product.
Video
How it works
User scenario
- Opens the stream and sees the host already in character
- Gets the show format through the chosen persona
- Watches the broadcast without the feeling of a "laggy filter"
Team / admin scenario
- Picks a character / persona for the broadcast
- Starts the camera and the stream with live face swap
- Monitors live stability (fps / drops)
Technology
- Mode
- Live face swap · ~60 fps
- Feature
- Character / persona selection
- NDA
- No public brand
Timeline: Production tool under NDA
Result and business effect
Implemented live face swap for streams: on air instead of the camera face - a chosen character or persona, change in realtime. Mechanics demo in the “Video” block (start on click). Brand and public link closed by NDA - without revealing the client or commercial metrics.
A fit if you need
Frequently asked questions
Questions about this type of project, the same ones clients ask before a similar build.
Why not build it on a no-code platform?
For testing a hypothesis over a couple of months, a builder is fine. Custom code pays off once you hit limits on permissions, integrations, speed or plan quotas. A common path: prototype on a builder, then an MVP on your own code with the data migrated.
How much does it cost and how do I pay?
We price per task, with an estimate in 1–2 days after the brief. It depends on the roles, screens and integrations in the release. You pay per iteration — prototype, MVP, next stage — on acceptance, not the whole budget upfront.
Who owns the code?
You do. The code and infrastructure stay with your company, so you can switch teams without rewriting the product.
What if the requirements change along the way?
Before coding we agree what goes into v1 and what goes to the backlog. New ideas do not break the current stage: we estimate them separately and plan them for the next one. Every finished task goes to staging, so you see the product instead of waiting for the end.
Need custom AI for live streaming?
We’ll discuss privately: broadcast scenario, characters, fps, and NDA constraints.
B2C NDA
AI tutor for kids
AI tutor MVP: homework, progress control, and a question bank by subject.
KhML
KhML - Explanatory Note XML
MVP: a set of construction documents → OCR/AI fields → design documentation + SIG → valid explanatory note XML. Stage 1 of 4.
FAVORIT
Web SaaS for marketplace cards
Stage 2: realtime generation, premium UI, infocard and model modes.