A family of Zyphra models for synthesizing speech and transferring voice features from a sound example. For ZONOS3-Mini, a local start is described.
One phrase in several variants
ZONOS accepts text and can target a voice sample. For the ZONOS3-Mini version, the authors describe the adjustment of pace, emotion, and pitch. Start with a short phrase that has a pause and one word with an important accent. It is easier to separate the quality of pronunciation from the similarity of the timbre. If you change both the original voice and the parameters of the pitch, it is difficult to understand where the difference came from.
Sound Route for Audio Guide
Imagine an audio guide for a small exhibition. Write down the voice of the person who allowed it to be used as a sample. Prepare a neutral fragment without music and echo, then synthesize the invitation to go to the next hall. Compare several options at the same volume. If the timbre is similar, but the model swallows the name of the hall, change the spelling of the word or break a long sentence into two phrases.
Listen to the next piece along with the first. For a consistent audio guide, the stability of the voice from one clue to another is important. Watch for the length of pauses between phrases.
Launch options and voice rights
ZONOS3-Mini materials have a native way to run and a WebGPU option. Suitable mode depends on the device and the required speed. The conditions of a specific version and cloud startup may vary, so check the model card before preparing a large series. Permission to record voice and permission to use it for synthesis is better to stipulate separately, especially if the result will be published.
Suitable tasks
- Prepare several options for your own reading
- Verify voice for instructional audio instruction
Strengths
- Synthesis of text with a voice guide
- Feed Options and Local Launch Options
Limitations
- Support for languages and settings depends on the version
- The voice pattern requires the permission of the speaker
Selection and conditions
- Free access
- Not confirmed
- Trial period
- Not confirmed
- No sign-up required
- Not confirmed
- Russian interface
- Not confirmed
- Local installation
- Yes
- API
- Not confirmed
“Not confirmed” means there is not enough information. The feature may still be available; check the developer’s terms.
- Input formats
- Text, Audio
- Result
- Audio
Terms
Zyphra publishes examples and launchers. For ZONOS3-Mini, there are local options, including the WebGPU environment, if there is a compatible device.From task to first result
Practical example
- Task
- Prepare several options for your own reading
- What to prepare
- Own voice recording with synthesis resolution and several new phrases in a suitable language.
Task brief or prompt
Speak phrases on a voice pattern. Check the names, intonation and naturalness of pauses.
Expected result
Black audio track with synthesized speech.
How to check the result
- Verify the text with the entire recording.
- Compare the timbre on different phrases and check the voice permission.
An editorial teaching example of a task brief, not the result of hands-on testing of the service. Check that the required mode is available, that you have rights to the source material and that the intended use of the result is permitted.
Check details
Official sources
- Zyphra's publication about ZONOS3-Mini and launch options
- Official ZONOS3-Mini card
- ZONOS2 card and technical materials
This description is based on the developer’s material. Features, prices and availability may change; check the primary source before use.
