AI models ranked by task
Public human votes, one task at a time, next to our own tests of how far each model goes.
As of Oct 19, 2026, the public-vote leaders by task are Wan 3.0 for text-to-video, MiniMax H3 Max for image-to-video, Seedance 2.5 for reference-to-video, GPT Image 2.5 Sunburst for text-to-image, and GPT Image 2.5 Sunburst for image editing.
- Updated Oct 19, 2026
- Methodology v3.3 →
Speech quality rankings
Independent human listening votes for text-to-speech. These Elo scores do not enter the Spicy Index.
Observed Sep 29, 2026 · 5 catalog models matched against a source leaderboard of 92 models.
- 1Gemini 3.8 Flash TTS1,267 Elo95% CI 1,251–1,283
2,211 listening votes
- 2Gemini 3.8 Flash-Lite TTS1,241 Elo95% CI 1,225–1,257
2,180 listening votes
- 3MiniMax Speech 2.8 HD1,171 Elo95% CI 1,160–1,182
4,498 listening votes
- 4MiniMax Speech 2.8 Turbo1,151 Elo95% CI 1,140–1,162
4,391 listening votes
- 5Chatterbox1,023 Elo95% CI 1,012–1,034
4,603 listening votes
Ranks use point estimates. Overlapping 95% confidence intervals do not establish a clear quality difference.
Artificial Analysis
Short English speech transcription
Six fixed read-speech clips from six speakers: 53.6 seconds and 146 reference words. This small diagnostic does not measure lyrics, noisy speech or overall transcription ability, and does not enter the Spicy Index.
- 1.37% WER
Lower is better
- HeartMuLa Transcribe
Updated Sep 30, 2026
4 substitutions · 0 deletions · 2 insertions / 146 reference words
4.11% WERLower is better
Word error rate (WER) is total substitutions, deletions and insertions divided by all reference words. Text is normalized consistently for case and punctuation; internal apostrophes are kept. Only complete six-clip results are shown. WER can exceed 100%.
Transcription methodBest text-to-video models
Text to video · 18 models ranked
Top pick rated Uncensored: Wan 3.0 · Freedom Score 96.0
See the full rankingBest image-to-video models
Image to video · 21 models ranked
Top pick rated Uncensored: MiniMax H3 · Freedom Score 70.7
See the full rankingBest reference-to-video models
Reference to video · 8 models ranked
Top pick rated Uncensored: Seedance 2.5 · Freedom Score 83.6
See the full rankingBest text-to-image models
Text to image · 24 models ranked
Top pick rated Uncensored: Qwen Image 3.0 · Freedom Score 96.0
See the full rankingBest AI image editing models
Image editing · 20 models ranked
Top pick rated Uncensored: Seedream 5.0 Pro · Freedom Score 76.9
See the full ranking