Bernini R2V delivers high-quality video from references, no speech

User reports that Bernini R2V produces video from reference images with significantly higher fidelity than LTX 2.3, but cannot generate speech dialogues. The model appears to excel at preserving subject consistency.