350M streamlined Turbo architecture
According to the model card, Turbo uses a streamlined 350M-parameter architecture intended to need less compute and VRAM than prior models.
Open Source Model Profile · ResembleAI
chatterbox-turbo
Chatterbox-Turbo is a ResembleAI text-to-speech model described as a streamlined 350M-parameter Turbo variant. It is recorded with MIT licensing and voice-cloning tags.
- Publisher
- ResembleAI
- Task
- text-to-speech
- Model type
- Unknown
- License
- mit
- Library
- Unknown
- Publication status
- Accepted · not indexed
Model overview
Chatterbox-Turbo is published by ResembleAI as a text-to-speech model in the Chatterbox family. The card introduces it as the most efficient model yet on a streamlined 350M-parameter architecture. Hub tags mark speech, speech-generation, voice-cloning, and English, with MIT licensing.
Recorded capabilities
Distilled one-step decoder
The card states the speech-token-to-mel decoder was distilled from 10 steps to one while retaining high-fidelity output.
Native paralinguistic tags
The publisher states paralinguistic tags are now native to the Turbo model.
Use cases in the source record
- Text-to-speech synthesis with voice cloning by supplying an audio prompt, as shown in the card’s generate-with-audio-prompt example.
- Multilingual synthesis experiments matching the card’s Chinese-text example with a language identifier.
Limitations and unknowns
- No Safetensors parameter count, architecture class, or model-type record was extracted; the 350M figure is a publisher claim.
- No evaluation results were extracted from this record.
- Latency, VRAM, and quality comparisons in the card are publisher descriptions and were not independently measured.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
Source and provenance
Source: ResembleAI/chatterbox-turbo
Captured: Unknown. Processed: 2026-09-07T19:34:36.310845+00:00.
Chatterbox TTS Made with ❤️ by Chatterbox is a family of three state-of-the-art, open-source text-to-speech models by Resemble AI. We are excited to introduce Chatterbox-Turbo , our most efficient model yet. Built on a streamlined 350M parameter architecture, Turbo delivers high-quality speech with less compute and VRAM than our previous models. We have also distilled the speech-token-to-mel decoder, previously a bottleneck, reducing generation from 10 steps to just one , while retaining high-fidelity audio output. Paralinguistic tags are now native to the Turbo model, allowing you to use [cough] , [laugh] , [chuckle] , and more to…
F001F002F003F004F005F006F007F008F009F011F012