The creative model reference

AI lip sync model comparison

Review the lip sync board for speech-driven video tasks and talking characters.

September 2026 benchmark snapshot · Curated by Oakgen. Data source and limits.

video / Lip Sync

The scores, in perspective.

5 models · 95% intervals

Swipe to compare all models →

Oakgen Arena · v1.0

11,123

Seedance 2.5

21,060

Seedance 2.0

31,054

Seedance 2.0 Mini

41,000

MiniMax H3Anchor

5923

Wan 3.0

↥ 95% confidence interval · Anchor interval not reported.5 ranked models

Relative scores on a zero-based display scale. Compare within this board only. Values are rounded for display.

September 2026 benchmark snapshot · Curated by Oakgen. Data source and limits.

Published reference / v1.0

Lip Sync leaderboard

Lip Sync video rankings · Benchmark index 1.0
Rank	Model	Overall score	95% confidence interval	Vote appearances	Distinct judges
01	Seedance 2.5ByteDance Criterion scores Audio quality1,178.195% CI 1,128.4 to 1,228.7Reused fit from master-video Consistency1,194.795% CI 1,151.8 to 1,243.3Reused fit from master-video Multi-character1,369.995% CI 1,308 to 1,459.9 Multilingual1,10595% CI 1,032.8 to 1,172.2 Physics & motion1,103.695% CI 1,068.6 to 1,137.4Reused fit from master-video Prompt adherence1,052.695% CI 1,014.6 to 1,088.4Reused fit from master-video Singing1,01295% CI 871.7 to 1,198.2 Subjective aesthetics1,096.495% CI 1,061.3 to 1,131.1Reused fit from master-video Talking1,003.995% CI 920.3 to 1,070.8	1,123.4LEAD	1,088.3 to 1,165.9	4,379	Not reported
02	Seedance 2.0ByteDance Criterion scores Audio quality1,102.895% CI 1,063.5 to 1,140.8Reused fit from master-video Consistency1,060.895% CI 1,022.7 to 1,099Reused fit from master-video Multi-character1,191.995% CI 1,107.5 to 1,304.6 Multilingual1,082.895% CI 1,016.9 to 1,157.5 Physics & motion1,036.195% CI 1,002.3 to 1,069.3Reused fit from master-video Prompt adherence1,050.595% CI 1,017.4 to 1,084.1Reused fit from master-video Singing1,01295% CI 875.5 to 1,222.2 Subjective aesthetics971.895% CI 935.7 to 1,010.3Reused fit from master-video Talking981.995% CI 912 to 1,044.9	1,060.3#2	1,027.8 to 1,104	4,423	Not reported
03	Seedance 2.0 MiniByteDance Criterion scores Audio quality1,127.895% CI 1,085.5 to 1,172.8Reused fit from master-video Consistency1,03495% CI 997.3 to 1,069.8Reused fit from master-video Multi-character1,102.195% CI 959.2 to 1,287.7 Multilingual1,139.195% CI 1,067.2 to 1,217.4 Physics & motion1,024.495% CI 991.6 to 1,059.6Reused fit from master-video Prompt adherence971.495% CI 937.1 to 1,002.7Reused fit from master-video Singing997.395% CI 849.6 to 1,175.8 Subjective aesthetics1,008.595% CI 972.9 to 1,048.3Reused fit from master-video Talking1,01495% CI 950.7 to 1,072.5	1,054.2#3	1,017.3 to 1,104.5	4,422	Not reported
04	MiniMax H3MiniMax Criterion scores Audio quality1,00095% CI Anchor · interval not reportedReused fit from master-video Consistency1,00095% CI Anchor · interval not reportedReused fit from master-video Multi-character1,00095% CI Anchor · interval not reported Multilingual1,00095% CI Anchor · interval not reported Physics & motion1,00095% CI Anchor · interval not reportedReused fit from master-video Prompt adherence1,00095% CI Anchor · interval not reportedReused fit from master-video Singing1,00095% CI Anchor · interval not reported Subjective aesthetics1,00095% CI Anchor · interval not reportedReused fit from master-video Talking1,00095% CI Anchor · interval not reported	1,000Anchor	Anchor · interval not reported	4,415	Not reported
05	Wan 3.0Alibaba Criterion scores Audio quality1,059.695% CI 1,022.9 to 1,100.1Reused fit from master-video Consistency1,121.595% CI 1,080.6 to 1,165.8Reused fit from master-video Multi-character925.895% CI 814.5 to 1,052 Multilingual896.895% CI 828.5 to 955.9 Physics & motion1,026.795% CI 993 to 1,058.4Reused fit from master-video Prompt adherence980.895% CI 940.2 to 1,021.2Reused fit from master-video Singing813.195% CI 717.2 to 951.9 Subjective aesthetics1,044.395% CI 1,006.5 to 1,083.1Reused fit from master-video Talking844.995% CI 766.9 to 914.1	923.1	893.3 to 957.8	4,372	Not reported

Ranks are the source’s published positions. Vote appearances are not unique people and must not be added across boards. Scores are rounded for display.

September 2026 benchmark snapshot · Curated by Oakgen. Data source and limits.

From comparison to creation

Choose for your brief.

Use speech with pauses, fast phrases and varied mouth shapes. Watch for timing drift and changes to facial identity. Test the language, camera angle and audio quality you will actually use rather than relying on a single demo.

Explore Oakgen's model directory

A source listing does not confirm current availability in Oakgen.

How this board is composed

Criteria and weights

Multi-character17.5%
Multilingual17.5%
Physics & motion6%
Consistency6%
Talking17.5%
Audio quality6%
Singing17.5%
Prompt adherence6%
Subjective aesthetics6%

Read the methodology