Can the next clip finish before playback ends?
Compare RTR only when resolution, duration, audio, endpoint, and timing scope are visible.
TASK-SPECIFIC COMPARISONS
A faster model is not automatically the right stream, avatar, or world. Comparisons are published only when both sides share meaningful fields.
COMPARE THE RIGHT CLOCK
Compare RTR only when resolution, duration, audio, endpoint, and timing scope are visible.
Compare time to first frame, cadence, jitter, stalls, reconnection, and whether prompts can change mid-stream.
Compare action-to-effect latency, supported controls, FPS, session horizon, and state consistency.
CLASS A · CLIP GENERATION
Quality-focused launch model or newer higher-throughput endpoint? Compare access, timing evidence, prompt expansion, and known gaps.
Independent runs are being prepared for streaming and world model comparisons. We will not publish a winner from incompatible vendor claims.