Enhancing Motion Variation in Text-to-Motion Models via Pose and Video Conditioned Editing