--- title: "Chatterbox Nano vs Turbo vs Multilingual V3: Mac Setup" description: "Nano is 110M, Turbo 350M, and Multilingual V3 500M. Compare languages, CPU vs MPS setup, and hear recorded Turbo audio before choosing." canonical: "https://www.murmurtts.com/blog/chatterbox-nano-turbo-v3-comparison" --- Comparison # Chatterbox Nano vs Turbo vs Multilingual V3: Mac Setup and Audio Choose a Chatterbox model for your Mac, hear recorded Turbo examples, and distinguish Nano, Turbo, and Multilingual V3 setup and support. ![](https://www.murmurtts.com/murmur-icon.webp)Murmur·August 12, 2026·5 min read Start with Nano for English speech on a CPU, Turbo for expressive English cloning, or Multilingual V3 for cross-language speech. Resemble AI lists them at 110M, 350M, and 500M parameters respectively. The choice also depends on the runtime you use: upstream Python examples and the models available inside Murmur are separate options. **What you can hear here** The recordings below demonstrate Chatterbox Turbo. This guide does not include Nano or V3 recordings, a same-script comparison, or measured Mac speed and memory results. Upstream setup examples were checked on 9 September 2026. ## Quick decision table | Model | Official size | Language scope | Best first use | | --- | --- | --- | --- | | Nano | 110M | English | CPU and constrained on-device inference | | Turbo | 350M | English | Low-latency expressive cloning and agents | | Multilingual V3 | 500M | 23+ | Cross-language voice cloning and localization | | Single Language Pack | 500M each | Six dedicated models | Language-specific quality control | | Original Chatterbox | 500M | English | Creative CFG and exaggeration control | ## Listen to Chatterbox Turbo These recorded examples are also available in [Murmur’s sample library](https://www.murmurtts.com/samples). Listen for the reactions and the words around them. Each clip uses its own script, so use them to assess delivery rather than compare model speed or accuracy. Chatterbox Turbo · Comedy · English0:00 **Script excerpt:** Okay so I tried that viral thirty day cold shower challenge and honestly? `[laugh]` The first week was absolute torture… Chatterbox Turbo · Suspense · English0:00 **Script excerpt:** `[sigh]` The cabin had been empty for eleven years… `[gasp]` A single set of fresh footprints leading to the basement stairs. ## Choose a Mac setup path | Model | Upstream setup | In Murmur | | --- | --- | --- | | Nano | The Python Turbo class with nano=True; CPU is documented | Not in the current model catalog | | Turbo | The Python Turbo class; use the official setup instructions | English voice cloning through an optional model download | | Multilingual V3 | The multilingual class with t3_model="v3"; CPU and MPS are documented | Listed as Chatterbox Multilingual; V3 checkpoint identity is not confirmed | For the native app route, open Murmur’s model setup, install Chatterbox Turbo, and choose a clean reference clip in Clone Voice. Turbo is English-only. For multilingual speech, select Chatterbox Multilingual and a supported language. Do not assume that a model named “Multilingual” uses the upstream V3 checkpoint. For a Python setup, the [official project](https://github.com/resemble-ai/chatterbox#installation) documents Python 3.11 and was developed on Debian. Use an isolated environment on your Mac, install `chatterbox-tts`, and follow the example for the exact variant. These are documentation-based starting points, not installation results measured for this article. `python3.11 -m venv .venv` `source .venv/bin/activate` `python -m pip install chatterbox-tts` Nano uses `ChatterboxTurboTTS.from_pretrained(device="cpu", nano=True)`. Multilingual V3 uses `ChatterboxMultilingualTTS.from_pretrained(device="mps", t3_model="v3")`; CPU is another documented option. Leaving out `t3_model="v3"` selects the legacy multilingual checkpoint in the current upstream example. Continue with the [Chatterbox V3 Mac setup guide](https://www.murmurtts.com/blog/chatterbox-v3-apple-silicon-setup) for the full multilingual workflow. ## Nano is a deployment decision The official Chatterbox table describes Nano as a 110M English model with the same broad architecture direction as Turbo, paralinguistic tags, and CPU performance of about three times real time on eight cores. That makes it interesting for assistants, games, previews, and accessibility tools where a GPU should remain available. The claim is an upstream reference point, not a Mac benchmark. Measure first audio, total real-time factor, and CPU use on the minimum device. Nano's small size does not mean it is the correct audiobook narrator. Long-form stability, pronunciation, and acting range need separate tests. It may be ideal as a responsive draft voice while a larger model produces final lines. ## Turbo is for expressive low-latency English Turbo reduces the decoder path from ten steps to one according to Resemble AI and supports reaction tags such as `[laugh]`. It targets English agents and production where response time and expressive events matter. A zero-shot clone still depends heavily on clean reference audio and can reproduce room noise or performance artifacts. Test tags in context. A laugh that sounds good alone may break timing in a final edit. Measure the spoken words around the event, the length of the inserted reaction, and whether repeated generations drift. Murmur includes Chatterbox in a broader [local model decision guide](https://www.murmurtts.com/blog/best-local-tts-models-2026) rather than presenting tags as universal quality. ## Multilingual V3 changes the evaluation V3 is a 500M general-purpose multilingual model with 23-plus languages. Resemble AI says it improves speaker similarity, reduces hallucinations, and produces more natural multilingual speech compared with earlier versions. Verify all three claims on a fixed local evaluation. Cross-language cloning needs native-speaker review because a voice can remain recognizable while stress, names, and borrowed words sound wrong. Do not assume every MLX conversion labeled Chatterbox is V3. Check the upstream commit or model card used for the conversion. Store that identity with results. For English and German methodology, see the [bilingual voice-cloning scorecard](https://www.murmurtts.com/blog/bilingual-voice-cloning-english-german). ## Choose by production stage A useful workflow can use more than one model. Nano can produce immediate previews. Turbo can generate an expressive English host. Multilingual V3 can preserve a speaker across localization. A language pack can handle a priority market. The production system should show which model generated each clip and should not silently substitute one family without updating saved metadata. Compare the same 250-word passage, clone reference, sample rate, and export settings. Score exact completion, critical tokens, speaker similarity, native pronunciation, first audio, warm speed, memory, and license. Hear Murmur's current output on the [samples page](https://www.murmurtts.com/samples) before downloading every checkpoint. ## Compare a passage from your own project Use one 250-word English passage and a clean reference recording from a speaker who has agreed to the test. Run it through the models you are considering with the same output settings. For V3, add a passage in your target language and ask a fluent speaker to review it. Follow the [local TTS testing guide](https://www.murmurtts.com/blog/local-tts-blind-benchmark-script-completion) to count skipped words, repeated phrases, and changes to names or numbers. - Save the exact model revision, runtime, voice reference, and settings with each output. - Time the first run separately from repeat runs with the model loaded. - Keep failed and incomplete recordings in the comparison. - Listen for reaction tags that interrupt the spoken words or change the timing. ## Practical checklist 1. Pick English CPU, English expressive, multilingual, or language-specific intent. 2. Verify the exact upstream version behind any conversion. 3. Use the same consented reference and script across models. 4. Score completion and pronunciation before expressiveness. 5. Measure first audio, warm speed, peak memory, and CPU or GPU use. 6. Save model family and version beside each clip. ## Frequently asked questions ## Sources - [Resemble AI Chatterbox official repository](https://github.com/resemble-ai/chatterbox)Accessed 2026-09-09 - [Chatterbox official demos](https://resemble-ai.github.io/chatterbox_demopage/)Accessed 2026-08-08 - [Chatterbox models on Hugging Face](https://huggingface.co/ResembleAI)Accessed 2026-08-08 ## Turn the research into finished local audio Murmur brings local models, reusable voices, projects, queueing, and export into one Mac production workspace. Hear the current models before choosing a workflow. [Buy Murmur · $49](https://murmur-licenses.tarunyadav9761.workers.dev/checkout) macOS 15+ · Apple Silicon required · 7-day refund policy