EmergingMultimodal and generative world models

Generative Video Researcher

Also called Video Generation Scientist

A Generative Video Researcher builds models that produce video that stays consistent over time and responds to control. They make video generation something a company can ship and trust, not just demo.

What they do

The work centres on training and improving video generation models. That means designing diffusion and transformer architectures that hold objects, motion and lighting steady across frames. The researcher runs experiments on conditioning, so a user can steer a clip with text, images, sound or action inputs. They study failures such as flicker, drifting identity and broken physics, then change data, architecture or training to fix them. They also build the evaluation side: automated media evaluators, plus human review that judges motion and continuity. Some work with simulators and digital twins to test control and physical plausibility. Much of the day is reading outputs closely and deciding what to try next.

What a strong hire delivers

In the first months, a strong hire gets a working video model baseline running on the team's training stack and shows measured gains against it. They set up an evaluation package that covers temporal consistency, controllability and artefact rate, with both automated and human review. They name the main failure modes, rank them by impact, and show which fixes worked. The team ends up with a clearer research direction and results it can trust.

How you can tell they are good

Ask them to walk through a video model they trained or improved and show the failures, not only the best clips. A strong candidate explains why a artefact appeared and what changed it. They can describe how they measured temporal consistency and why automated scores alone were not enough. Listen for honest accounts of dead ends. Be wary of anyone who talks only about image quality or cherry-picked samples.

Why it is hard to fill

Few people have trained video models end to end, and most sit inside frontier labs or large media research groups. The title is rarely used. The work often hides under multimodal researcher, diffusion researcher or simulation scientist. Candidates come from mixed backgrounds, so a title search misses many of them. Compute access matters too, because people who have not had it may show strong ideas but thin evidence. We read papers, code and project detail to tell the difference.

Where to find them

Look at multimodal ML researchers, generative modelling scientists and simulation or graphics researchers who have moved towards video. Strong people also come from world model and robotics teams, where action-conditioned prediction is familiar. They work at frontier labs, creative software companies, film and media technology groups, and university labs with good compute. Open research code and published video work are useful signals of who has done the training themselves.

Indicative pay

Not enough published evidence yet to give a fair range for this role, which is common for work this new. We benchmark it for you as part of a search.

Interview questions for a Generative Video Researcher

  1. Tell us about a video model you trained. What was the hardest failure to fix?

    A strong answer names a specific failure such as flicker or identity drift, explains the diagnosis, and states what changed and by how much it helped. Worry if they only describe successes or credit the result to scale alone.

  2. How do you measure temporal consistency, and where do automated metrics mislead you?

    They should mention several measures, plus human review for motion and continuity. They should know metrics can reward blurry or static clips. Worry if they rely on one score or treat human evaluation as optional.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We are building generative video models and need a researcher to make them consistent, controllable and reliable. You will train and improve models, design how we judge them, and work closely with engineers who take the results into the product. We value shown work over titles.

What you will do

  • Train and improve video generation models using diffusion and transformer methods
  • Reduce flicker, identity drift and other artefacts across longer clips
  • Add and test control through text, image, sound and action inputs
  • Build automated and human evaluation for temporal consistency and artefact rate
  • Run clear experiments and report results, including failures
  • Use simulators and digital twins to test control and physical plausibility
  • Work with engineers to move research into production

What you will bring

  • Demonstrated work training generative video or closely related models
  • Strong grounding in diffusion and temporal modelling
  • Experience with transformer-based multimodal modelling
  • Ability to design evaluations that go beyond a single metric
  • Clear written explanations of experiments and results
  • Comfort working on large multimodal training stacks

Nice to have

  • Experience with simulation or digital twins
  • Published or open research on video or media generation
  • Experience running human evaluation studies

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Skills that matter

  • Multimodal modellingAdvancedCombine information across modalities
  • Generative mediaAdvancedProduce controllable visual, sound or spatial outputs
  • Transformer modellingWorkingImplement attention-based sequence or multimodal models
  • Simulation and digital twinsWorkingBuild useful virtual representations of environments

Other roles in multimodal and generative world models

Hiring nowMultimodal Research ScientistA Multimodal Research Scientist builds models that connect text, images, sound and action, and designs the benchmarks that show where they break…EmergingWorld Model ResearcherA World Model Researcher builds models that predict how an environment will change, given what has happened and what an agent does next. That gives…EmergingGenerative 3D ResearcherA Generative 3D Researcher builds models and pipelines that produce spatial assets with controllable structure. They matter because generated…

Related roles in other families

Hiring nowPretraining Research EngineerFoundation model pretrainingA pretraining research engineer runs controlled experiments on how large models learn from broad data. The work decides which architectures, data…Hiring nowModel Architecture ResearcherFoundation model pretrainingA Model Architecture Researcher tests changes to how a model is built, to improve its behaviour or efficiency. The work tells a company which…EmergingScaling Science ResearcherFoundation model pretrainingA Scaling Science Researcher works out how model performance changes with compute, data and model size. Their estimates tell a company how to spend…EmergingTraining Dynamics ResearcherFoundation model pretrainingA Training Dynamics Researcher works out why large training runs wobble, stall or diverge, and fixes the cause. Fewer failed runs means less wasted…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert.

← Multimodal and generative world modelsAll new job families

Start a search

Hiring a Generative Video Researcher?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.