ElevenLabs vs Fish Audio
| Measure | ElevenLabsEleven v4 | Fish AudioFish Audio Drama 3 |
|---|---|---|
| Score (of 100) | 93 (better) | 72 |
| Rank overall | #1 (better) | #9 |
| Boards listed on | 3 of 3 | 2 of 3 |
| Price per 1M characters | $80.00 | $15.00 (better) |
| Coval time to first audio (p50) | — | — |
| Speko time to first audio (p50) | 844 ms | — |
| Humanness Index time to first audio (median) | 671 ms | — |
| Open weights | No | No |
"—" means the source has not published a figure for that model. Latency sources are never averaged together.
Short answer
Eleven v4 scores higher (93 to 72), but Fish Audio Drama 3 costs 5.3× less. The latency sources split on who is faster.
Quality
Eleven v4 (ElevenLabs) scores 93, 21 points clear of Fish Audio Drama 3 (Fish Audio) at 72. They rank #1 and #9 of 119 models.
Fish Audio Drama 3 is listed on only 2 of 3 boards, so its score has less behind it.
Price
Fish Audio Drama 3 lists at $15.00 per 1M characters. Eleven v4 lists at $80.00, 5.3× as much.
ElevenLabs also has Eleven v4 Turbo at $40.00, which scores 77.
Speed
Coval has no figure for Eleven v4, so ElevenLabs is represented by Eleven v4 Turbo; Coval has no figure for Fish Audio Drama 3, so Fish Audio is represented by S2.1 Pro. Coval: Eleven v4 Turbo at 190 ms, S2.1 Pro at 360 ms. ElevenLabs is 1.9× faster.
Speko has no figure for Fish Audio Drama 3, so Fish Audio is represented by S2.1 Pro. Speko: S2.1 Pro at 185 ms, Eleven v4 at 844 ms. Fish Audio is 4.6× faster.
Humanness Index has no figure for Fish Audio Drama 3, so Fish Audio is represented by S2.1 Pro. Humanness Index: S2.1 Pro at 141 ms, Eleven v4 at 671 ms. Fish Audio is 4.8× faster.
The sources do not agree on who is faster. They use different harnesses, regions and sample sizes, so treat speed as unsettled.
Model lineups
| Model | Rank | Score | Price$ / 1M chars | Covalp50 ms | Spekop50 ms | Humanness Indexmedian ms |
|---|---|---|---|---|---|---|
| Eleven v4ElevenLabs | #1 | 93 | $80.00 | — | 844 ms | 671 ms |
| Eleven v4 TurboElevenLabs | #6 | 77 | $40.00 | 190 ms | 205 ms | 158 ms |
| Eleven v3ElevenLabs | #8 | 74 | $100.00 | — | — | 758 ms |
| v3 ConversationalElevenLabs | #16 | 66 | $50.00 | — | 266 ms | — |
| Turbo v2.5ElevenLabs | #50 | 37 | $50.00 | — | — | — |
| Fish Audio Drama 3Fish Audio | #9 | 72 | $15.00 | — | — | — |
| Fish Audio Drama 3 PreviewFish Audio | #19 | 61 | $15.00 | — | — | — |
| Fish Audio S2.1 Pro (Oct 2026)Fish Audio | #26 | 54 | $15.00 | — | — | — |
| Fish Audio S2 ProFish Audio | #42 | 42 | $15.00 | — | — | — |
| OpenAudio S1Fish Audio | #54 | 34 | $15.00 | — | — | — |
ElevenLabs has 3 more models on the provider page. Fish Audio has 2 more models on its page.
ElevenLabs has 8 models on the board, ranging from 93 down to 9; 4 sit in the top 20. Fish Audio has 7 models, from 72 to 16; 2 in the top 20.
Open weights: Fish Audio S2 Pro (Fish Audio, #42). The other side lists none.
Questions people ask
- Is Fish Audio better than ElevenLabs?
- On our score, no. Eleven v4 (ElevenLabs) scores 93, 21 points clear of Fish Audio Drama 3 (Fish Audio) at 72. They rank #1 and #9 of 119 models. Fish Audio Drama 3 lists at $15.00 per 1M characters. Eleven v4 lists at $80.00, 5.3× as much.
- Which is cheaper, ElevenLabs or Fish Audio?
- Fish Audio Drama 3 lists at $15.00 per 1M characters. Eleven v4 lists at $80.00, 5.3× as much. ElevenLabs also has Eleven v4 Turbo at $40.00, which scores 77.
- Which is faster, ElevenLabs or Fish Audio?
- Coval has no figure for Eleven v4, so ElevenLabs is represented by Eleven v4 Turbo; Coval has no figure for Fish Audio Drama 3, so Fish Audio is represented by S2.1 Pro. Coval: Eleven v4 Turbo at 190 ms, S2.1 Pro at 360 ms. ElevenLabs is 1.9× faster. Speko has no figure for Fish Audio Drama 3, so Fish Audio is represented by S2.1 Pro. Speko: S2.1 Pro at 185 ms, Eleven v4 at 844 ms. Fish Audio is 4.6× faster. Humanness Index has no figure for Fish Audio Drama 3, so Fish Audio is represented by S2.1 Pro. Humanness Index: S2.1 Pro at 141 ms, Eleven v4 at 671 ms. Fish Audio is 4.8× faster. The sources do not agree on who is faster. They use different harnesses, regions and sample sizes, so treat speed as unsettled.
Where the numbers come from
Quality is the 0–100 score from our rank aggregation of 3 public boards: artificialanalysis.ai (Speech Arena Elo), voicearena.com (arena Elo, US English) and humannessindex.vapi.ai (humanness score 0–100). Prices are dollars per 1M characters, taken from Artificial Analysis and provider pages; per-minute prices are converted to the same unit. Latency is time to first audio (the wait before sound starts), copied from Coval, Speko and Humanness Index and never averaged across them. We do not measure anything ourselves. How latency is sourced.