Model · Realtime video
Realtime Video
A video generation model that produces each frame before the previous one has finished showing, so it can be steered while it runs.
Model card
- Identifier
- truth-rtv
- Area
- Realtime video
- Status
- Active
- Version
- none released
- Access
- Planned tier for 0.1 is open weights and code.
- Licence
- To be stated at release.
Model names start with truth- after the foundation's mission. The name is not a claim that the model's outputs are correct; each model card states what has been measured.
Intended use
- Research on interactive and long-form video generation.
- Education and live production, through products, under their usage policies.
Out of scope
- Generating the likeness of a real person without their consent.
- Content designed to deceive viewers about what is real.
Status now
The architecture, data policy and evaluation suite are being defined. There is no public model and nothing measured yet; this card describes the plan.
Data
- Sources
- Licensed and public-domain footage only.
- Licences
- The data recipe, with the licence of every source, is published with each release.
Evaluation
Planned
- Temporal consistency, with three metrics and a statement of what each misses.
- Prompt adherence.
- Latency per frame on stated hardware, with timing code published.
What comes next
| When | What |
|---|---|
| Q4 2026 | Evaluation suite and data policy published |
| Q1 2027 | truth-rtv-0.1: first weights, offline generation |
| Q2 2027 | truth-rtv-0.2: realtime generation on a single GPU |