Update docs/source/en/using-diffusers/consisid.md

Co-authored-by: Steven Liu <[email protected]>

Author: SHYuanBest

docs/source/en/using-diffusers/consisid.md

@@ -14,7 +14,7 @@ specific language governing permissions and limitations under the License.
 [ConsisID](https://github.com/PKU-YuanGroup/ConsisID) is an identity-preserving text-to-video generation model that keeps the face consistent in the generated video by frequency decomposition. The main features of ConsisID are:
 
 - Frequency decomposition: The characteristics of the DiT architecture are analyzed from the frequency domain perspective, and based on these characteristics, a reasonable control information injection method is designed.
-Consistency training strategy: A coarse-to-fine training strategy, dynamic masking loss, and dynamic cross-face loss further enhance the model's generalization ability and identity preservation performance.
+- Consistency training strategy: A coarse-to-fine training strategy, dynamic masking loss, and dynamic cross-face loss further enhance the model's generalization ability and identity preservation performance.
 - Inference without finetuning: Previous methods required case-by-case finetuning of the input ID before inference, leading to significant time and computational costs. In contrast, ConsisID is tuning-free.
 
 This guide will walk you through using ConsisID for use cases.