Research Scientist in Interactive Avatars and Video Diffusion at Synthesia (Remote in Europe)
Job Description
This video diffusion research scientist job is with Synthesia’s Interactive Avatars team. You will work on avatar video models that can hold a real-time conversation. The role is remote within Europe.
At a glance
- Position: Research Scientist, Interactive Avatars, full time
- Where: Remote in Europe; offices in London, Munich and Zurich
- Team: Interactive Avatars, part of an R&D department of more than 40 people
- Pay: Not stated in the posting
- Posted: 6 October 2026
- Apply: open until filled (re-checked 14 November 2026)
About the interactive avatar research
- Shape research on two-person (dyadic) interaction and own clear problems from start to finish
- Improve how an avatar perceives the user’s audio and video and reacts to it
- Post-train multimodal models to produce natural two-way conversations
- Adapt diffusion models to new signals, such as turn-taking and listener cues
- Build ways to measure the quality of an interaction
- Plan data needs with the data team, run careful experiments and share results
Why this role matters
Most video generation today is one-way. Real-time avatars that listen and respond are a new research problem. It mixes video diffusion, audio and conversation modelling, and few teams work on it at this scale.
What Synthesia is looking for
- A strong machine learning background and hands-on work with diffusion models, ideally for video or avatars
- Papers at top venues (CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR or SIGGRAPH) on world models, dyadic interaction or video diffusion, or similar impact
- Experience taking research through to working code
- PyTorch and large-scale training tools
- Clear communication (the posting does not state a degree requirement)
Nice to have
- Real-time or streaming generation, including autoregressive video diffusion
- Distillation for low-latency inference
- Audio-driven face, gesture or body motion models
- Conversation modelling, such as turn-taking or backchannels
- Mentoring experience
Pay and location
The posting gives no salary figure. The role is remote within Europe, and Synthesia has offices in London, Munich and Zurich. The posting does not say which countries you can be employed from, so ask early if you live outside these three.
How to apply
Apply through the official Synthesia job posting. Synthesia’s posting does not give a closing date, so the role is open until filled; ResearchJobs.in will re-check this listing on 14 November 2026. Check the posting for location and work-authorisation details before you apply.
See all our artificial intelligence jobs, or browse more research jobs on ResearchJobs.in.
Hiring institution: Synthesia
Official advertisement: jobs.ashbyhq.com
How to prepare for this application
- Show one video or demo of your generative model work; a short clip says more than a long description
- Revise video diffusion basics: temporal attention, latent video models and conditioning on audio
- Read about turn-taking and backchannels in conversation research so you can talk about listener behaviour
- Think about how you would measure a good interaction, both with people and with automatic metrics
- Be ready to discuss latency: distillation, fewer sampling steps and streaming generation
About Synthesia
Synthesia is an AI video company founded in 2017 and headquartered in London. It makes a platform that businesses use to create videos with AI avatars. The posting says it has raised more than US$530 million in total funding.
