V
Train a custom voice
Prepare a voice corpus, transcribe it, and fine-tune F5-TTS through the guided desktop workflow and CLI.
Local voice and face production
Turn a script into a repeatable video workflow.
Your voice. Your face. Your production stack.
Turn a script into a repeatable voice-and-face production workflow. Train and generate locally, then connect Agent Avatar to Agent Studio when the finished video needs a full creative pipeline.
Qualified technical beta — Linux or WSL2 with an NVIDIA GPU. Request access for the current build.
Agent Avatar / local pipelinebeta
./avatar generate \ --voice my-voice \ --face my-face \ --text "Here's what shipped this week." \ -o out/update.mp4 # batch a script through the local API curl localhost:8765/lipsync/generate/batch \ -d @release-notes.json
Local
training and generation workflow
3
voice, video, and photo engines
API
scriptable local production surface
Beta
qualified access for GPU users
See it in motion
The suite overview today. Avatar walkthroughs next.
Start with the Agent Software Suite overview, then follow along as we publish videos made with Agent Avatar about Agent Avatar.
Agent Software Suite overview
A short introduction to the wider suite. Dedicated Agent Avatar walkthroughs will be added as they are produced.
Output first
Build the presenter layer for the videos you already need to make.
Agent Avatar is a standalone production engine for creators and technical teams. Use it directly, or connect it to Agent Studio when the result needs editing, motion, music, review, and final export.
F
Lip sync from footage
Use MuseTalk to drive a source video with generated speech for repeatable talking-head scenes.
P
Animate a photo
Use the photo-based workflow when a headshot is the right starting point instead of a source video.
B
Render whole scripts
Batch a sequence of scenes and keep word-level timing available for captions and downstream editing.
E
Tune the finish
Apply optional face enhancement and choose how much polish fits the source and the story.
A
Use a local API or CLI
Drive supported workflows from the unified avatar CLI or authenticated local HTTP endpoints.
L
Catalog voices and faces
Name and reuse voice, face, and source entries instead of managing fragile file paths by hand.
S
Connect to Agent Studio
When you own both products, Agent Studio can use Avatar capabilities inside a broader editing and production workflow.
How it works
Capture once. Compose repeatedly.
1
Capture with consent
Record the voice, footage, or photo you are authorized to use and prepare the source material for training or generation.
2
Train locally
Run the voice and face preparation pipeline on a supported NVIDIA GPU, with progress, checkpoints, and diagnostics.
3
Write and render
Submit a line or a batch script through the desktop app, CLI, or local API and produce synchronized output.
4
Finish in Studio
Connect Agent Avatar to Agent Studio when the result needs timeline editing, motion, music, review, and final export.
Voice training and TTS (F5-TTS)
Lip sync (MuseTalk)
Photo animation (SadTalker)
Optional face enhancement
Voice, face, and source catalog
Local HTTP API and unified CLI
Interface layer
Everything above is reachable over a local HTTP API, the unified ./avatar CLI, and the Agent Software Suite plugin interface.
Local by design
Keep sensitive processing close to the machine.
Local training
Voice and face preparation runs on a supported local GPU workflow.
Clear boundaries
Source media and model files are not uploaded for processing; entitlement, updates, and optional integrations may use network services.
Consent first
Train and generate only with your own likeness or explicit, informed permission from the subject.
The boundary
Avatar is standalone. Studio makes it part of the full production.
Agent Avatar
Train and run voice, face, lip-sync, photo animation, enhancement, catalog, CLI, and local API workflows.
Agent Avatar + Agent Studio
When you have both products, Studio can unlock supported Avatar-powered scenes inside timeline editing, motion, music, review, and export workflows.
Explore Agent StudioComparison
The difference is where your processing happens.
Local processing path
Agent AvatarYesHeyGenNoSynthesiaNoD-IDNo
Runs on your own GPU
Agent AvatarYesHeyGenNoSynthesiaNoD-IDNo
Offline generation after setup
Agent AvatarYesHeyGenNoSynthesiaNoD-IDNo
Scriptable local API
Agent AvatarYesHeyGenYesSynthesiaYesD-IDYes
Batch a whole script
Agent AvatarYesHeyGenYesSynthesiaYesD-IDNo
Agent Studio connector
Agent AvatarYesHeyGenNoSynthesiaNoD-IDNo
| Capability | Agent Avatar | HeyGen | Synthesia | D-ID |
|---|---|---|---|---|
| Local processing path | Yes | No | No | No |
| Runs on your own GPU | Yes | No | No | No |
| Offline generation after setup | Yes | No | No | No |
| Scriptable local API | Yes | Yes | Yes | Yes |
| Batch a whole script | Yes | Yes | Yes | No |
| Agent Studio connector | Yes | No | No | No |
Consent policy
Clone yourself. Or someone who said yes.
You may train a voice or face model on yourself, or on a person who has given explicit, informed permission. Do not impersonate people, mislead an audience about synthetic media, or use someone's likeness to obtain money, credentials, or trust.
Read the full consent policyFAQ
Common questions
Who is Agent Avatar for?
It is for creators, educators, developers, and technical teams who need repeatable voice-and-face video and can run a local GPU workflow. Nontechnical self-serve onboarding is not the current target.
What do I need to run it?
The current beta targets Linux or WSL2 with an NVIDIA GPU offering at least 8GB of VRAM, about 32GB of system RAM, and roughly 100GB of storage for models and media.
What does local mean here?
Training and generation are designed to run on your machine. Source media and model files are not uploaded for processing. Account entitlement, updates, and optional connected services may still use network services.
Can I use Agent Avatar with Agent Studio?
Yes. Agent Avatar is standalone, and users who have both products can unlock supported Avatar-powered voice, face, and lip-sync workflows inside Agent Studio.
Can I clone anyone?
Only yourself or a person who has given explicit, informed consent. Impersonation and deceptive synthetic media are prohibited. Read the consent policy before training or generating.
Is public pricing available?
Not yet. The product is in a qualified technical beta while packaging, onboarding, and production commerce are finalized. Request access for the current availability and support path.
Make the next video without filming the next take.
Join the technical beta, or talk to us about a finished Launch Reel for your product.