List your company

ResearchFormerly WizWand

MEAD

Benchmarks

TaskDatasetSOTA resultTrendResults

Speech-Preserving Facial Expression Manipulation

MEAD (intra-identity)

0.572FAD

48

Speech-Preserving Facial Expression Manipulation

MEAD (cross-identity)

1.872FAD

38

Talking Head Generation

MEAD

72Realism Score

32

Audio-Driven Facial Animation

MEAD 41 (test)

31.873PSNR

26

Emotional Talking Head Generation

MEAD Cross-ID

45.2Score (Neutral)

24

Self-reenactment portrait animation

MEAD 59 (test)

0.952CSIM

18

Audio-driven talking head generation

MEAD

8.778Sync

14

Emotional Talking Head Generation

MEAD Intra-ID

9.452Fidelity: Neutral

12

Portrait Image Animation

MEAD (test)

32.696FID

12

Talking Head Generation

MEAD self-driven

26.36FID

10

Talking Face Generation

MEAD Neutral

8.29LSE-C

10

3D Face Reconstruction

MEAD

58.09R Eye Error

9

Audio-driven Video Generation

MEAD

3.77FID

8

Talking Head Generation

MEAD

45.403FID

8

Dynamic Facial Expression Recognition

MEAD 8-class

88.44WAR

8

Talking Head Generation

MEAD

41.81FID

7

3D GAN Inversion

MEAD (novel views)

0.223LPIPS (±60°)

7

Audio-driven talking head synthesis

Mead 60 (test)

1.76LSE-C

7

Talking Head Generation

MEAD

4.16Avg-R

6

Talking Face Generation

MEAD

0FID

6

3D Talking Face Synthesis

Emotional talking face dataset (MEAD) M003 and M030 (test)

8.163Sync-C

6

Emotional Talking Head Generation

MEAD

4.65Emotion Score

6

Emotion-conditioned generation

MEAD

0.653E-Score

5

Talking Head Generation

MEAD User Study

4.33Visual Quality Score

5

Facial Component Control

MEAD 53 (test)

13.695Pose Error

5