Updated November 3, 2026. The leaderboard turned over again since our September edition. ElevenLabs Eleven v4 (September 28) took back first place on the independent Artificial Analysis TTS arena, with its Turbo variant a close second, at a lower price than v3 used to cost. Google Gemini 3.8 Flash TTS entered the top five at an introductory price of about $0.0135 per minute. Fish Audio Drama 3 is the best-ranked open-weight model. For live voice agents, Phonic-1 (October 29) and xAI Grok Voice 3.0 now sit above OpenAI GPT-Live-1 and Gemini 3.8 Live.
This guide is for founders and marketers who need voiceovers, podcast intros, ads, and explainer narration, plus teams building a voice agent. Picks come first, then the ranking, prices, and how to choose.
Need a voiceover today, not a vendor evaluation? Paste your script into Teamday Audio and get a downloadable voiceover from Gemini TTS, kept next to the video and files it belongs to.
Create your first voiceover →Quick Answer: The Best AI Voice Model for Each Job
| Your job | Start with | Why |
|---|---|---|
| Ad, explainer, or brand voiceover | ElevenLabs Eleven v4 | #1 on the TTS arena, 90+ languages, voice cloning, full editor; $0.08 per 1K characters |
| Voice agent that needs a top-ranked voice | Eleven v4 Turbo or Cartesia Sonic 3.6 | #2 and #4 on the arena; ~100 ms median inference (ElevenLabs), under 90 ms claim (Cartesia) |
| Best quality per dollar | Gemini 3.8 Flash TTS | #5 on the arena at about $0.0135 per audio minute until December 31, 2026 |
| Cheap high-volume narration | Gemini 3.8 Flash-Lite TTS or Deepgram Aura-2 | About $0.009 per audio minute (introductory) / $0.030 per 1K characters |
| Live voice agent on the phone or web | Phonic-1, Grok Voice 3.0, or GPT-Live-1 | Top three on the speech-to-speech index; $0.07, $0.08, and $0.05 per minute |
| Self-hosted, small business | Fish Audio Drama 3 | Highest-ranked open weights (#6); commercial use only under $500K annual revenue |
| Self-hosted, no license strings | Qwen3-TTS or Kokoro-82M | Apache 2.0 weights; Qwen3-TTS clones from 3 seconds of audio |
Do not choose from a 20-second demo. Test your own script, your product names, and every target language. For a downloadable-model deployment, see our Hugging Face TTS guide with CPU audio samples; for recognizing recorded speech, see our transcription guide with local test results.
November 2026 Ranking: Text-to-Speech Quality
Artificial Analysis runs blind A/B listening tests and converts the votes into ELO scores. It is the best public quality signal, but it measures short clips, not your brand voice, pronunciation of your product names, or long-form consistency. Several neighbors are within a few points of each other, which is inside the arena's confidence intervals: treat ranks 1–2 and 8–9 as ties.
| Rank | Model | Vendor | Arena ELO | Released | List price |
|---|---|---|---|---|---|
| 1 | Eleven v4 | ElevenLabs | 1334 | Sep 2026 | $0.08 / 1K chars |
| 2 | Eleven v4 Turbo | ElevenLabs | 1331 | Sep 2026 | $0.04 / 1K chars |
| 3 | Qwen-Audio-3.1-TTS-Plus | Alibaba | 1301 | Sep 2026 | AA lists $19.30 / 1M chars |
| 4 | Sonic 3.6 | Cartesia | 1293 | Aug 2026 | Plans from $5/mo (100K credits) |
| 5 | Gemini 3.8 Flash TTS | 1282 | Sep 2026 | $0.50 text in / $9 audio out per 1M tokens (intro) | |
| 6 | Drama 3 | Fish Audio | 1272 | Oct 2026 | Open weights; AA lists $15 / 1M chars via API |
| 7 | Lightning V3.1 Pro | Smallest.ai | 1269 | Oct 2026 | AA lists $19.50 / 1M chars |
| 8 | Qwen-Audio-3.0-TTS-Plus | Alibaba | 1266 | Jul 2026 | AA lists $19.30 / 1M chars |
| 9 | Realtime TTS-2 | Inworld | 1266 | Aug 2026 | $25 / 1M chars on demand |
| 10 | Gemini 3.8 Flash-Lite TTS | 1254 | Sep 2026 | $0.50 text in / $6 audio out per 1M tokens (intro) | |
| 15 | Voxtral 3 TTS | Mistral | 1218 | Oct 2026 | AA lists $13 / 1M chars |
| 17 | Gemini 3.1 Flash TTS | 1213 | Apr 2026 | $1 text in / $20 audio out per 1M tokens | |
| 18 | v3 Conversational | ElevenLabs | 1212 | Aug 2026 | $0.04 / 1K chars |
Source: Artificial Analysis TTS leaderboard, checked November 3, 2026. Ranks 11–14 and 16 are Speechify Simba 3.2, Fish Audio Drama 3 Preview, BreezeBlue Breeze TTS 2, AssemblyAI-TTS-1, and Inworld Realtime TTS-2 Flash. Vendor prices are from each vendor's pricing page on the same date. Where a vendor does not publish a per-character rate we could confirm, we show Artificial Analysis's normalized price and label it "AA lists".
Realtime Voice Agents: Speech-to-Speech Ranking
| Rank | Model | Index score | Price |
|---|---|---|---|
| 1 | Phonic-1 (High) | 87.1 | $0.07/min Starter, $0.065/min Pro ($200 monthly minimum); no separate LLM charge |
| 2 | xAI Grok Voice 3.0 | 86.9 | $0.08/min |
| 3 | OpenAI GPT-Live-1 (Astra, medium) | 83.2 | $0.05/min, billed per second, plus the backend reasoning model |
| 4 | Gemini 3.8 Live Extended Thinking (High) | 82.6 | $0.005/min audio in, $0.018/min audio out |
Source: Artificial Analysis speech-to-speech, Phonic pricing, xAI pricing, OpenAI pricing, Gemini API pricing, all checked November 3, 2026. ElevenLabs' Eleven Realtime v1 Max is #7 (77.3) and OpenAI GPT-Realtime 2.1 High #8 (75.2).
Phonic-1 is a single end-to-end model rather than a transcribe-reason-speak pipeline, trained on 50+ languages; Phonic claims 600 ms responses. It is the newest entrant here and has the shortest production track record, so run your own call scripts before switching. xAI's Grok Voice 2.0 is now marked deprecated in favor of 3.0 at the same price. GPT-Live-1 is a full-duplex voice layer that hands reasoning and tool calls to a backend model you pick; the older GPT-Realtime 2.1 is still offered at $32 per 1M audio input tokens and $64 per 1M audio output tokens, and its mini version at $10 and $20.
Price Per Finished Minute
Most TTS vendors price per character. A spoken minute at a normal pace of about 150 words is roughly 900 characters. The table below converts list prices on that assumption. It is an estimate, not a vendor quote. Google publishes the Gemini 3.8 TTS rates per 10 seconds of audio, so those rows are Google's own numbers.
| Model | List price | ≈ Per spoken minute |
|---|---|---|
| MiniMax Speech 2.8 HD | $100 / 1M chars | ~$0.09 |
| ElevenLabs Eleven v4, v3, Multilingual v2 | $0.08 / 1K chars | ~$0.072 |
| ElevenLabs Eleven v4 Turbo, v3 Conversational, Flash | $0.04 / 1K chars | ~$0.036 |
| Deepgram Aura-2 | $0.030 / 1K chars | ~$0.027 |
| Inworld Realtime TTS-2 | $25 / 1M chars | ~$0.023 |
| Inworld Realtime TTS-2 Flash | $15 / 1M chars | ~$0.0135 |
| OpenAI tts-1, xAI TTS | $15 / 1M chars | ~$0.0135 |
| Gemini 3.8 Flash TTS | $0.00225 per 10 s of audio until Dec 31, 2026; $0.0045 from Jan 1, 2027 | $0.0135, then $0.027 |
| Gemini 3.8 Flash-Lite TTS | $0.0015 per 10 s until Dec 31, 2026; $0.003 from Jan 1, 2027 | $0.009, then $0.018 |
| gpt-4o-mini-tts, Gemini 3.1 Flash TTS | Priced per audio token | Measure with a real script |
Gemini 3.8 TTS prices double on January 1, 2027. If you are budgeting a 2027 content calendar, use the second number. The generation fee is rarely what drives the cost anyway. Retakes, pronunciation fixes, and review time usually cost more. Track total spend divided by approved finished minutes, not the list price.
The Models, One by One
ElevenLabs Eleven v4: best voiceover quality and the most complete product
Eleven v4 (September 28, 2026) is ElevenLabs' new flagship and #1 on the arena. It covers 90+ languages, adds Professional Voice Clones on top of Instant Voice Clones from 10 seconds of audio, and is available in the ElevenLabs creative app, its agents product, and the API. Eleven v4 Turbo is the low-latency version for agents, with ElevenLabs reporting about 100 ms median inference and 150 ms to first speech.
The API price list also dropped: v4, v3, and Multilingual v2 are now $0.08 per 1K characters (v3 and Multilingual v2 were $0.10 in September), and v4 Turbo, v3 Conversational, and Flash are $0.04. In September we wrote that ElevenLabs no longer won blind votes; v4 reverses that. It is again the default for a marketer who wants a voice library, cloning, an editor, dubbing, and music under one account.
ElevenLabs Flash v2.5, stock voice "Sarah". Generated through the API on July 13, 2026. This is the older Flash model, not v4.
Cartesia Sonic 3.6: strongest independent pick for streaming
Sonic 3.6 (August 27, 2026) dropped from #1 to #4, but its score rose and it is still the top model from a vendor other than ElevenLabs or a hyperscaler. Cartesia claims under 90 ms latency and covers 44 languages. Pricing is credit-based: Pro is $5/month for 100K credits and includes instant voice cloning, Startup is $49 for 1.25M with two professional clone slots, and Scale is $299 for 8M (pricing).
Google Gemini 3.8 Flash TTS and Flash-Lite TTS: best quality per dollar
Google released Gemini 3.8 Flash TTS and Flash-Lite TTS in late September. Flash ranks #5 and Flash-Lite #10, both above Gemini 3.1 Flash TTS (#17), which is still labeled preview. At introductory prices of $9 and $6 per 1M audio output tokens through December 31, 2026, they are among the cheapest voices in the arena's top ten (Flash-Lite is the cheapest); prices double on January 1, 2027 (pricing). Both have a free tier in AI Studio. On Google's paid API tier, prompts and outputs are not used to improve Google's products (Gemini API terms).
Gemini 3.8 Live is Google's realtime conversation model. It is now #4 on the speech-to-speech index and is the cheapest of the top four at $0.005 per minute of audio in and $0.018 per minute out.
Alibaba Qwen-Audio-3.1 and Inworld TTS-2
Alibaba shipped the Qwen-Audio-3.1 series on September 23 and cut prices across its voice models. Qwen-Audio-3.1-TTS-Plus is #3 on the arena. Model Studio prices the Flash version per token ($0.23 in / $1.87 out per 1M tokens in Singapore, model page), so measure a real script before comparing it with per-character vendors. It is API-only, not open weights.
Inworld's Realtime TTS-2 is $25 per 1M characters on demand, dropping as low as $5 on Enterprise, and TTS-2 Flash is $15. TTS-2 fell from #3 to #9 as newer models arrived, but it remains one of the cheaper options in the top ten.
OpenAI: GPT-Live-1, GPT-Realtime 2.1, gpt-4o-mini-tts
- GPT-Live-1: $0.05/min plus the reasoning model; #3 on speech-to-speech.
- gpt-4o-mini-tts ($0.60 per 1M text tokens in, $12 per 1M audio tokens out), tts-1 ($15 per 1M characters), and tts-1-hd ($30): OpenAI's plain TTS options.
We found no newer OpenAI standalone TTS model and no OpenAI TTS model on the arena's top 20 as of November 3 (pricing).
MiniMax Speech 2.8, Mistral Voxtral 3, Deepgram Aura-2
- MiniMax Speech 2.8: HD at $100 per 1M characters, Turbo at $60. Rapid voice cloning costs $1.50 per voice, charged on first use (pricing). Not in the arena's top 20.
- Mistral Voxtral TTS 3 (
voxtral-mini-tts-3-0, October 2026): preset voices plus zero-shot voice cloning, 34 emotion tags, and a separate streaming model for live conversations (docs). API-only; no weights were released for version 3. - Deepgram Aura-2: $0.030 per 1K characters, and $0.075/min as a full voice agent (pricing).
Deepgram Aura-2, stock voice "Thalia". Generated through the API on July 13, 2026, from the same script as the ElevenLabs sample.
Open weights: Fish Audio Drama 3, Qwen3-TTS, Kokoro, Chatterbox, Voxtral
| Model | License | Notes |
|---|---|---|
| Fish Audio Drama 3 | Fish Audio Community License | 35B MoE (3B active), 83 languages, cloning, multi-speaker; #6 on the arena. Commercial use only below $500K annual revenue, no hosted TTS service, "Built with Fish Audio" attribution required |
| Qwen3-TTS 0.6B / 1.7B | Apache 2.0 | 10 languages, 3-second voice cloning |
| Kokoro-82M | Apache 2.0 | Tiny, fast, 8 languages, no cloning |
| Chatterbox Multilingual v3 (Resemble) | MIT | 25 languages, cloning, watermark on all output |
| Voxtral-4B-TTS (Mistral) | CC BY-NC 4.0 | Non-commercial weights; Voxtral TTS 3 is API-only |
Drama 3 is the first open-weight model to rank alongside the commercial leaders, but it needs a data-center GPU (Fish Audio's reference numbers are on an H200), and its license is not open source in the usual sense: read Section III before you build a product on it. For laptop-scale models with audio samples, see the Hugging Face TTS guide.
PlayHT is gone: the platform shut down on December 31, 2025, after Meta hired the team, according to migration notices from competitors; we found no primary Meta announcement. Sesame has no public API. Microsoft MAI-Voice-2 and Amazon Nova 2 Sonic are options for procurement fit inside Azure or AWS; neither appears in either Artificial Analysis top list we checked.
The voice is one piece of the job. In Teamday you write the script with an AI teammate, generate the voiceover in Audio, add music, and keep every version with the rest of the campaign, with no separate accounts to juggle.
Open Teamday Audio →How to Choose a Voice Model for Your Business
- Name the job. Recorded voiceover (TTS) and live conversation (speech-to-speech) are different products with different leaders. Pick from the right table above.
- Run your own script. Test 60–90 seconds with your product names, numbers, URLs, and prices, in every language you ship. Have a native speaker judge each language.
- Check the rights for your plan. Commercial use normally needs a paid plan. For a cloned voice, keep a written consent record from the speaker. For open weights, read the license: Drama 3 has a revenue cap and Voxtral-4B-TTS is non-commercial.
- Price per approved minute, for the whole year. Count retakes and fixes, not only the list price, and check for introductory rates that expire (Gemini 3.8 TTS doubles on January 1, 2027).
- Keep provenance. Store the provider, model ID, voice ID, script version, generation date, and approver with every published file. When a model is retired or a customer asks what they heard, you can answer and regenerate.
Which Voice Models Can You Use in Teamday?
| Where in Teamday | Models | Access |
|---|---|---|
| Audio studio (script → voiceover file) | Gemini 3.1 Flash TTS (preview), six preset voices | Your connected Google Gemini account |
| Voice conversations with agents | OpenAI GPT-Realtime 2.1; ElevenLabs Flash v2.5, v3, Multilingual v2 with Scribe v2 transcription; Deepgram Aura-2 with Nova-3 transcription | Your own OpenAI, ElevenLabs, or Deepgram key |
| Video narration in Reel's pipeline | ElevenLabs Flash v2.5 | Your own ElevenLabs key |
The Audio studio offers six Gemini voices (Kore, Puck, Charon, Aoede, Achird, Sulafat); add delivery directions such as "warm, unhurried" to the script. The same studio also makes background music with Google Lyria. Voiceovers are not paid from Teamday AI credits yet; they use your own Gemini account. Teamday does not yet offer Eleven v4, Cartesia, Gemini 3.8 TTS or Live, GPT-Live-1, Phonic-1, or voice cloning. Use those directly with the vendor if you need them now.
Frequently Asked Questions
What is the best AI voice model in November 2026?
On the Artificial Analysis TTS arena (blind listener votes, checked November 3, 2026), ElevenLabs Eleven v4 ranks first, just ahead of its low-latency sibling Eleven v4 Turbo, followed by Alibaba Qwen-Audio-3.1-TTS-Plus and Cartesia Sonic 3.6. For live voice agents, Phonic-1 and xAI Grok Voice 3.0 lead the speech-to-speech index, ahead of OpenAI GPT-Live-1 and Gemini 3.8 Live.
How much does AI voiceover cost per minute?
Roughly $0.01 to $0.09 per finished minute at API list prices, assuming about 900 characters per spoken minute. Eleven v4 and v3 are $0.08 per 1,000 characters (about $0.07 per minute), Eleven v4 Turbo and Flash $0.04 per 1,000, Deepgram Aura-2 $0.030 per 1,000, Inworld TTS-2 $25 per million characters, and Gemini 3.8 Flash TTS about $0.0135 per minute of audio at its introductory price through December 31, 2026. Retakes and edits usually cost more than the generation itself.
What is the difference between TTS and realtime speech-to-speech?
Text-to-speech reads a script you supply and is what you need for voiceovers, podcasts, and ads. Realtime speech-to-speech listens, reasons, and speaks in a live conversation, so interruption handling and turn latency matter as much as the voice. Phonic-1, Grok Voice 3.0, GPT-Live-1 and Gemini 3.8 Live are speech-to-speech; Eleven v4, Cartesia Sonic and Gemini TTS are TTS.
Can I use AI voices commercially?
Usually yes on paid plans, but check the provider's terms for your plan, the specific stock voice, and consent for any cloned voice. Open-weight models differ: Kokoro-82M and Qwen3-TTS are Apache 2.0; Fish Audio Drama 3 allows commercial use only below $500,000 in annual revenue and never as a hosted TTS service; Mistral's Voxtral-4B-TTS weights are CC BY-NC 4.0 (non-commercial).
Which voice models can I use in Teamday?
Teamday's Audio studio turns scripts into voiceovers with Google Gemini 3.1 Flash TTS and six preset voices, using your connected Gemini account. Voice conversations with agents support OpenAI GPT-Realtime 2.1, ElevenLabs Flash v2.5, v3 and Multilingual v2, and Deepgram Aura-2 with your own provider keys. Video narration in Reel's pipeline uses ElevenLabs Flash v2.5.
Test it on your own script: one paragraph in, a voiceover file out, and you judge the result instead of a demo reel.
Generate a voiceover from your script →Related Guides
- Best AI music generators
- Best AI video models
- Best frontier AI models
- Best AI avatar models
- Reel, Teamday's video producer
- Teamday model catalog
Research cutoff: November 3, 2026. Rankings are from Artificial Analysis. Prices and model status come from each vendor's pricing page or documentation on that date. Teamday availability was checked against the current Teamday code. Prices and preview status change often, so confirm them before committing a production budget.