GamiWays

TTS & Voice Synthesis

Comparison of voice synthesis solutions for conversational pipelines (2025–2026). Benchmarks, strategic stakes, and decision questions.

🎯

Strategic framing: hear TTS in context

What the numbers say

TTFA measures audible start; ELO is a preference signal, not a guarantee for your script. Treat both as signals, then listen to short turns, pauses, numbers and sensitive lines.

What shapes the experience

Cloning, languages and especially performance direction determine how the voice speaks. Distinguish explicit control from simple instructions: test whether intent, pace and pauses hold from one turn to the next.

Stay alert to

Price per hour is an entry point, not a production budget: include volume, streams, caching and regeneration. Also check cloning consent and rights, data residency, and fallback if a voice or API changes.

Reading method: set the required control and voice level first, then compare latency and quality on the same lines, and finally test cost, rights and fallback across a complete session.

CLOUD APIs

This section compares cloud streaming TTS APIs (2025–2026). They speed up integration, with vendor dependency and sovereignty constraints to assess for each deployment context.

Click a header to sort

SolutionTTFAELOCloningPerformance directionMultilingualPrice/hr · entryDetail
Eleven v4 / v4 TurboUpd
Phase 1 MVP — Référence qualité
150 ms1329✓7/10 · Instruction-based direction✓ 90$2.40/hView →
Cartesia Sonic 3.6Upd
Phase 1 MVP — Latence critique
90 ms1285✓10/10 · Explicit API control✓ 44$2.25/hView →
Gemini 3.8 Flash TTSNew
Direction de jeu vocale — à tester en français
unpublished1275✓7/10 · Instruction-based direction✓ 130API price pendingView →
Inworld Realtime TTS-2 + FlashUpd
Phase 1 MVP — Qualité + Pipeline conversationnel complet + 100+ langues
100 ms1262✓7/10 · Instruction-based direction✓ 200$1.35/hView →
Gemini 3.8 Flash-Lite TTSNew
Cascade temps réel — à chiffrer
unpublished1248✓7/10 · Instruction-based direction✓ 101API price pendingView →
StepAudio 2.5 TTS
Benchmarking qualité — non recommandé pour production EU
200 ms1214✓7/10 · Instruction-based direction✓ 100$4.59/hView →
HeyGen Voice 1New
Candidat TTS — Où est Ava ?, sans remplacer Gradium
unpublished1196✓7/10 · Instruction-based direction✓ 39$1.80/hView →
Fish Audio S2.1 ProUpd
Phase 1 MVP — Coût/Souveraineté
200 ms1190✓7/10 · Instruction-based direction✓ 83$0.81/hView →
Gradium TTSNew
Phase B R&D — Lip-sync + Infrastructure Temps Réel
216 ms1174✓3/10 · Limited control✓ 5$2.60/hView →
MiniMax Speech 2.8New
Prototypage économique — non recommandé pour production EU
200 ms1173✓10/10 · Explicit API control✓ 17$3.24/hView →
OpenAI Realtime (gpt-realtime-2.1) + GPT-Live-1Upd
Dialogue live — délégation GPT-5.5
300 ms1079✗7/10 · Instruction-based direction✓ 70$4.61/hView →
Deepgram Flux TTSNew
Phase B R&D — Test conversationnel temps réel
80 ms1053✗10/10 · Explicit API control✗$2.70/hView →
Hume AI Octave 2Upd
Phase 1 MVP — Expressivité émotionnelle
100 ms1051✓10/10 · Explicit API control✓ 11API price pendingView →
Smallest.ai Lightning V3.1
Pipeline complète — Phase B Hydra S2S
100 ms1024✓3/10 · Limited control✓ 15$0.94/hView →
Gemini 3.8 LiveNew
Dialogue live — hors cascade TTS
unpublished—✗3/10 · Limited control✓ 97API price pendingView →
Gemini 3.8 Live Extended ThinkingNew
Dialogue live — raisonnement parallèle
unpublished—✗3/10 · Limited control✓ 97API price pendingView →
Fish Audio Drama 3New
Où est Ava ? — test de direction, hors antenne
unpublished—✓7/10 · Instruction-based direction✓ uncountedAPI price pendingView →
Deepgram Aura 2Upd
Phase 1 MVP — Stack ASR+TTS intégré
80 ms—✗3/10 · Limited control✗$1.62/hView →
xAI Grok TTSNew
Phase B R&D — Évaluation pipeline S2S
285 ms—✓7/10 · Instruction-based direction✓ 20$0.81/hView →

Compare by solution

Cloud APIUpdated

Eleven v4 / v4 Turbo

Eleven v4 (28 sept. 2026) — API $0.08/1k caractères, Turbo $0.04. Arena publique 17 oct. : v4 ELO 1329 (#2), Turbo 1330 (#1).

Instruction-based direction

Quality10/10
10
Latency8/10
8
Cloning10/10
10
Performance direction7/10
7
Sovereignty2/10
2
Pricing2/10
2
150 ms TTFAELO 1329CloningLip-sync
Phase 1 MVP — Référence qualité
Full details →
Cloud APIUpdated

Cartesia Sonic 3.6

Sonic 3.6 (27 août 2026) — 44 langues, plafond éditeur sous 90 ms. Arena publique 17 oct. : ELO 1285, rang 4.

Explicit API control

Quality10/10
10
Latency9/10
9
Cloning8/10
8
Performance direction10/10
10
Sovereignty4/10
4
Pricing5/10
5
90 ms TTFAELO 1285Cloning
Phase 1 MVP — Latence critique
Full details →
Cloud APINew

Gemini 3.8 Flash TTS

Nouveau — design de voix et acting ligne à ligne, 130 langues. Pas de TTFA publié.

Instruction-based direction

Quality8/10
8
Latency4/10
4
Cloning7/10
7
Performance direction7/10
7
Sovereignty2/10
2
Pricing5/10
5
unpublished TTFAELO 1275Cloning
Direction de jeu vocale — à tester en français
Full details →
Cloud APIUpdated

Inworld Realtime TTS-2 + Flash

TTS conversationnel disponible : direction libre, contexte audio, non-verbaux et variante Flash à faible délai serveur

Instruction-based direction

Quality9/10
9
Latency8/10
8
Cloning9/10
9
Performance direction7/10
7
Sovereignty6/10
6
Pricing7/10
7
100 ms TTFAELO 1262CloningSovereignLip-sync
Phase 1 MVP — Qualité + Pipeline conversationnel complet + 100+ langues
Full details →
Cloud APINew

Gemini 3.8 Flash-Lite TTS

Nouveau — variante débit et agents en cascade, 101 langues. Pas de TTFA publié.

Instruction-based direction

Quality7/10
7
Latency5/10
5
Cloning6/10
6
Performance direction7/10
7
Sovereignty2/10
2
Pricing6/10
6
unpublished TTFAELO 1248Cloning
Cascade temps réel — à chiffrer
Full details →
Cloud API

StepAudio 2.5 TTS

Contextual TTS — ELO AA V2 Free 1210 (5 octobre ; rang non publié), dual-level context control, zero-shot voice cloning in 3 sec

Instruction-based direction

Quality9/10
9
Latency6/10
6
Cloning9/10
9
Performance direction7/10
7
Sovereignty1/10
1
Pricing5/10
5
200 ms TTFAELO 1214Cloning
Benchmarking qualité — non recommandé pour production EU
Full details →
Cloud APINew

HeyGen Voice 1

Nouveau — TTS maison HeyGen, octobre 2026. Arena Controlled Voice, pas l’arena vocale principale.

Instruction-based direction

Quality9/10
9
Latency4/10
4
Cloning9/10
9
Performance direction7/10
7
Sovereignty1/10
1
Pricing6/10
6
unpublished TTFAELO 1196CloningLip-sync
Candidat TTS — Où est Ava ?, sans remplacer Gradium
Full details →
Cloud APIUpdated

Fish Audio S2.1 Pro

Pay-as-you-go voice cloning — 70% cheaper than ElevenLabs

Instruction-based direction

Quality6/10
6
Latency6/10
6
Cloning8/10
8
Performance direction7/10
7
Sovereignty4/10
4
Pricing7/10
7
200 ms TTFAELO 1190Cloning
Phase 1 MVP — Coût/Souveraineté
Full details →
Cloud APINew

Gradium TTS

TTS streaming français avec précision de prononciation structurée et résidence UE activable

Limited control

Quality7/10
7
Latency7/10
7
Cloning8/10
8
Performance direction3/10
3
Sovereignty6/10
6
Pricing9/10
9
216 ms TTFAELO 1174CloningSovereignLip-sync
Phase B R&D — Lip-sync + Infrastructure Temps Réel
Full details →
Cloud APINew

MiniMax Speech 2.8

Sound tags natifs — Speech 2.8 HD ELO 1173 (rang 22). Tarif d’entrée Turbo $60/1M, HD arena $100/1M.

Explicit API control

Quality8/10
8
Latency6/10
6
Cloning8/10
8
Performance direction10/10
10
Sovereignty1/10
1
Pricing10/10
10
200 ms TTFAELO 1173Cloning
Prototypage économique — non recommandé pour production EU
Full details →
Cloud APIUpdated

OpenAI Realtime (gpt-realtime-2.1) + GPT-Live-1

GPT-Live-1 parle, GPT-5.5 raisonne derrière. API au 10 sept. à 0,05 $/min pour la voix, plus le modèle. Scores The Batch contre AVM seulement.

Instruction-based direction

Quality4/10
4
Latency6/10
6
Cloning1/10
1
Performance direction7/10
7
Sovereignty2/10
2
Pricing3/10
3
300 ms TTFAELO 1079
Dialogue live — délégation GPT-5.5
Full details →
Cloud APINew

Deepgram Flux TTS

New — conversation-native TTS: 80ms first audio (vendor), cross-turn context, WebSocket + REST

Explicit API control

Quality7/10
7
Latency9/10
9
Cloning1/10
1
Performance direction10/10
10
Sovereignty6/10
6
Pricing4/10
4
80 ms TTFAELO 1053Sovereign
Phase B R&D — Test conversationnel temps réel
Full details →
Cloud APIUpdated

Hume AI Octave 2

Emotional TTS / speech-to-speech — API retirement announced for 13 Nov. 2026

Explicit API control

Quality3/10
3
Latency8/10
8
Cloning6/10
6
Performance direction10/10
10
Sovereignty2/10
2
Pricing8/10
8
100 ms TTFAELO 1051Cloning
Phase 1 MVP — Expressivité émotionnelle
Full details →
Cloud API

Smallest.ai Lightning V3.1

Full voice pipeline (TTS + STT + LLM + S2S) — sub-100ms, 15+ languages

Limited control

Quality2/10
2
Latency9/10
9
Cloning7/10
7
Performance direction3/10
3
Sovereignty6/10
6
Pricing8/10
8
100 ms TTFAELO 1024CloningSovereign
Pipeline complète — Phase B Hydra S2S
Full details →
Cloud APINew

Gemini 3.8 Live

Dialogue live (15 sept. 2026) — 97 langues, outils en arrière-plan. Pas un TTS de cascade. Pas de milliseconde publiée.

Limited control

Quality6/10
6
Latency4/10
4
Cloning1/10
1
Performance direction3/10
3
Sovereignty2/10
2
Pricing5/10
5
unpublished TTFA
Dialogue live — hors cascade TTS
Full details →
Cloud APINew

Gemini 3.8 Live Extended Thinking

Dialogue live à raisonnement parallèle (15 sept. 2026). Les indices sont des chiffres Google. Pas de milliseconde publiée.

Limited control

Quality7/10
7
Latency4/10
4
Cloning1/10
1
Performance direction3/10
3
Sovereignty2/10
2
Pricing5/10
5
unpublished TTFA
Dialogue live — raisonnement parallèle
Full details →
Cloud APINew

Fish Audio Drama 3

Preview — direct the performance in the script. No published price or time to first audio.

Instruction-based direction

Quality6/10
6
Latency4/10
4
Cloning7/10
7
Performance direction7/10
7
Sovereignty2/10
2
Pricing4/10
4
unpublished TTFACloning
Où est Ava ? — test de direction, hors antenne
Full details →
Cloud APIUpdated

Deepgram Aura 2

Aura 2 baseline — Coval: 327ms median TTFA, 5.1% WER (12 Aug 2026)

Limited control

Quality6/10
6
Latency9/10
9
Cloning1/10
1
Performance direction3/10
3
Sovereignty3/10
3
Pricing7/10
7
80 ms TTFA
Phase 1 MVP — Stack ASR+TTS intégré
Full details →
Cloud APINew

xAI Grok TTS

TTS naturel et expressif — #3 Humanness Index Vapi (93/100), 460ms TTFA, 5 voix, 20 langues, $15/1M chars

Instruction-based direction

Quality8/10
8
Latency8/10
8
Cloning7/10
7
Performance direction7/10
7
Sovereignty2/10
2
Pricing7/10
7
285 ms TTFACloning
Phase B R&D — Évaluation pipeline S2S
Full details →

Global architecture: Voice-to-Voice & real time

GPT-Live-1 and Gemini 3.8 Live are dialogue records in this table, not cascade TTS engines. Their architecture comparison — full duplex, delegation, vendor indexes — is on the Voice-to-Voice page. No unpublished millisecond figure is turned into a TTFA.

Explore Voice-to-Voice →