FishAudio
DomesticFree + PaidAudio Speech
Voice cloning, text-to-speech, emotional voiceover
Developer:FishAudio Team
FishAudioIntroduction
FishAudio is a popular open-source-friendly AI voice platform in China, specializing in few-shot voice cloning. With just a short reference audio clip, it can train high-fidelity custom voices. It also offers TTS text-to-speech capabilities, widely used for audiobooks, short video narration, virtual digital human voices, and podcast production. The platform supports long-text segmentation with precise control over pauses, emphasis, and speaking rate, delivering natural-sounding speech with minimal robotic artifacts. Users can manage all their trained voices online and generate audio on demand. A free tier allows users to experience voice cloning and text-to-speech, while paid compute packages unlock longer durations, higher concurrency, and lossless audio export. Open APIs enable seamless integration into custom projects, making FishAudio a top choice for creators and independent developers. The platform actively maintains its open-source models, ensuring high technical transparency.
FishAudio's Key Features
Key Product Features
- 1Few-shot voice cloning: train a custom voice with short reference audio
- 2Text-to-Speech (TTS) with SSML support for precise voice control
- 3Automatic long-text segmentation and synthesis, optimized for documents up to 10,000 words.
- 4Timbre Library Management: Save and Switch Custom Replicas
- 5Audio export: supports MP3 and WAV formats
- 6Open API with support for batch programmatic calls
FishAudio Pricing Plans
Free Tier, Paid Plans, Billing Options, Private Deployment & Enterprise Pricing
- 1Free quota: New users receive complimentary compute credits with daily usage limits.
- 2Paid: Compute Time Package
- 3Billing: Based on duration and compute power
- 4Private Deployment: Open-source version available for self-hosting; commercial private deployment options available upon request.
FishAudio User Guide
Entry URL, Registration Requirements, Brief Operation Steps, and API / Client Support
Entry Point:fish.audio
- 1Entry: https://fish.audio
- 2Sign Up: Register with email or phone number
- 3Steps: Log in to the platform and upload reference audio to train the voice.
- 4Paste text to read aloud, then select a voice
- 5Adjust parameters and preview
- 6Generate and download audio
- 7Client/API: Web, Open API
FishAudioParameter Overview
Information compiled from public sources; official information prevails.
- Is data used for training?
- Users can opt out of having their training data used for platform model iteration.
- Underlying model
- FishSpeech proprietary speech model
- Context length
- Supports segmented processing of long texts up to 10,000 characters.
- Multi-modal support
- Text, Audio
- Search Online
- No general web search available
- File Parsing
- mp3, wav, flac, m4a
- Enterprise Edition and Deployment
- No public information available
- Compliance Certification
- No public information available
- Domestic Availability
- No public information available
Field Description:Will my input be used for model training, and can I turn this off?
Field Description:Primary models supported by the product, or the range of switchable models
Field Description:The maximum context size per conversation determines the length of documents it can read.
Field Description:Supported input and output formats, including text, images, audio, video, and files
Field Description:Supports real-time web search and provides source links?
Field Description:Supports uploading files and parsing capabilities, such as PDFs, spreadsheets, and code.
Field Description:Does it offer an enterprise edition, on-premises deployment, and team permission management?
Field Description:Publicly disclosed security compliance certifications and generative AI service filing status
Field Description:Convenience of access and supported payment methods in mainland China