Choose theme
AI Resources
Fish Audio S2 Pro
Fish Audio S2 Pro turns text into speech, with delivery instructions placed inside the script.
This is a previous-generation Fish Audio model. The provider's current model guide recommends S2.1 Pro for production use. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
Text-to-speech model
Generate spoken audio from written lines.
Why it stands out
Instructions within the script
Bracketed descriptions can direct a particular part of the speech rather than setting one style for the whole clip.
Availability
Downloadable model and code
The S2 release includes model weights, fine-tuning code and a streaming inference engine.
Why it matters
What makes it useful
A narration line may need a pause in the middle, or a quieter delivery for one phrase. S2 accepts directions such as [pause] and [whisper] alongside the words to be spoken.
What to know
Where it fits
It sits at the speech-output stage of a narration or voice application. You supply the script; the model generates audio. The downloadable release and Fish Audio's hosted service are different access paths.
Notable points
What stands out
The controls use natural-language descriptions rather than a fixed menu of style names. A direction can be placed where the delivery should change, including within a sentence.
Before using
What to review
The model card identifies the Fish Audio Research License; follow its linked terms for the downloaded release.
Check the exact model offered by a hosted service before assuming it is this S2 Pro release.
Reader fit
Who may find it relevant
People experimenting with how a written line is spoken, especially when delivery changes within the line.
Builders maintaining an S2 workflow who need to distinguish its downloaded release from the provider's current model offering.
Editorial note
Why LifeHubber lists it
Fish Audio's guide distinguishes a sentence's mood from a change to one word: sentence-level emotion cues usually work best at the start, while [emphasis] goes just before the word or phrase to stress. That helps you choose where to put a direction when editing a spoken line.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
What to explore next
Compare the speech task you need.
S2 Pro generates speech from a script. The voice and meeting collection separates speech generation from transcription and voice-agent work.
More in Speech Models
Keep browsing this category
Explore more speech model resources.
KittenTTS
KittenML/KittenTTS
An ONNX-based text-to-speech library with small model variants, CPU support, built-in voices, and audio-file output.
AuK
Tencent-Hunyuan/AuK
A 1.5B speech model for instruction-guided text-to-speech, content and acoustic editing, paralinguistic changes, speech enhancement, and source separation, with public code, weights, demos, ComfyUI nodes, and fine-tuning materials.
PersonaPlex
NVIDIA/personaplex
A real-time full-duplex speech-to-speech conversational model with persona control through role prompts and voice conditioning.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
