Sesame CSM

Sesame CSM

A browser demo of the 1B Conversational Speech Model from Sesame, which generates speech from text and audio inputs. Built for research and education, with strict rules against impersonation and deception.

Made by sesame on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a browser demo of CSM, the Conversational Speech Model from Sesame, in its 1B size. The model takes text and audio as input and produces speech. Under the hood it pairs a Llama backbone with a smaller audio decoder. The authors released it for research and education.

Who it suits: It suits researchers, students and developers who want to hear a conversational speech model and see how it is built, and creators who want to try generated speech within the usage rules.

Industry | Who uses it | How they use it | Example
Research | Speech researchers | Listen to the model and study its design | Compare conversational and read prompts
Education | Students and teachers | Learn how speech generation models work | Play generated speech in a class
Software | Developers | Read the app and generator code | Open generator.py and models.py
Media | Creators | Try generated speech for drafts | Hear a short line spoken

You can skip it if you want to copy a real person's voice, since the authors forbid imitating real people without their consent.

Where to use it:

• A web browser: open the Space page on Hugging Face
• Windows, macOS or Linux: run a copy with Python

Use case | Who | What they get back
Generate speech | Creators and researchers | Audio made from your text and optional audio input
Study the model | Developers | Code for the generator and model
Use the sample prompts | Newcomers | Conversational and read-speech prompts that ship with the repository

How to set it up: The online page needs nothing. A local copy needs Python and the packages in requirements.txt.

• A web browser for the online demo
• Python for a local copy
• The packages listed in requirements.txt

1. To use it online, open the Space page on Hugging Face.
2. To run a copy, install the Hugging Face hub package.
3. Download the Space into a local folder.
4. Move into the folder and install Gradio 5.20.0 and the listed packages.
5. Start the app.

pip install -U huggingface_hub
hf download sesame/csm-1b --repo-type space --local-dir csm-1b
cd csm-1b
pip install gradio==5.20.0
pip install -r requirements.txt
python app.py

How to use it: Start the app from its folder, or open the Space online. Enter the text you want spoken, add audio context if the page offers it, and generate. The page then plays the result. The repository includes prompt audio to try.

python app.py

Example: You type one short conversational line and pick one of the sample prompts. The page returns audio of that line spoken in a natural, conversational style, and you can play it back.

For newcomers and experts: If you are new to this, try one short sentence with a sample prompt. Listening to the result is the best way to judge it.

If you know the field, the model outputs residual vector quantization audio codes from text and audio inputs. A Llama backbone feeds a smaller audio decoder, and the decoder produces codes for the Mimi audio codec. The Space names a Llama 3.2 1B model and the CSM 1B checkpoint. A file named watermarking.py sits in the repository.

You enter text and optional audio
A Llama backbone processes the inputs
A small decoder produces Mimi audio codes
The codes become audio you can play

Good to know: The project forbids three kinds of use, and the authors say they are not responsible for misuse. The sample prompts come from two public speech datasets, one with conversations and one with read speech.

• License: apache-2
• Cost to run: none named
• Data it reaches: the text and audio you give it, plus the sample prompts in the repository
• Not for: copying a real person's voice without consent, fake news or fraud calls, or anything illegal or harmful

On Hugging Face, 13 November 2026

• Likes: 868
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/sesame/csm-1b
• Files: https://huggingface.co/spaces/sesame/csm-1b/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.