Elftia integrates two categories of speech service — speech recognition (ASR, Automatic Speech Recognition) and speech synthesis (TTS, Text-to-Speech) — allowing you to switch freely between voice and text in conversations and content creation.
Use Cases
- Automatically transcribe voice messages or recorded audio files to text
- Convert AI replies or custom text to speech for playback
- Use voice input when typing is inconvenient
- Add narration to generated content
Speech Recognition (ASR)
The speech recognition feature transcribes speech from audio files into text.
Supported Providers
| Provider | Type identifier | Description |
|---|
| OpenAI Whisper | openai-whisper | OpenAI's Whisper speech recognition API; supports multiple languages |
| ElevenLabs Scribe | elevenlabs-stt | ElevenLabs' speech-to-text service |
| Fish Audio | fishaudio-asr | Fish Audio's speech recognition service |
Configuration Steps
- Open Settings > Media Providers > Speech Recognition
- Select an ASR provider
- Enter the API Key
- Select a default model (e.g.
whisper-1)
- Save and enable
- (Optional) Set it as the default provider
Configuration Parameters
| Setting | Description | Required |
|---|
| API Key | The provider's API key | Yes |
| Default model | The model used for transcription | No (uses provider default) |
| Default provider | Whether to set this as the default ASR provider | No |
How to Use
- Select or record an audio clip
- The system automatically sends the audio to the configured ASR provider
- The transcription result is returned as text
The ASR service has built-in automatic format conversion. If the uploaded audio file format is not natively supported by the Whisper API, the system attempts to automatically convert it to WAV using ffmpeg before transcription.
| Format | Extension |
|---|
| FLAC | .flac |
| M4A | .m4a |
| MP3 | .mp3 |
| MP4 | .mp4 |
| MPEG | .mpeg, .mpga |
| OGG | .oga, .ogg |
| WAV | .wav |
| WebM | .webm |
Other formats (such as .amr or .silk) are automatically converted to WAV before processing.
:::info ffmpeg Dependency
The automatic format conversion feature requires ffmpeg to be installed on the system. If ffmpeg is not installed, unsupported formats are sent directly to the API (and the API may refuse to process them).
:::
Transcription Parameters
| Parameter | Description |
|---|
providerId | Specifies the ASR provider to use (optional; defaults to the default provider) |
modelId | Specifies the model to use (optional) |
language | Specifies the audio language (optional; auto-detected if not set) |
Speech Synthesis (TTS)
The speech synthesis feature converts text content into natural-sounding speech.
Supported Providers
| Provider | Type identifier | Description |
|---|
| ElevenLabs | elevenlabs-tts | High-quality multilingual speech synthesis with a rich selection of voices |
| Fish Audio | fishaudio-tts | Fish Audio's speech synthesis service |
Configuration Steps
- Open Settings > Media Providers > Speech Synthesis
- Select a TTS provider
- Enter the API Key
- Select a default model
- Select a default voice
- Save and enable
- (Optional) Set it as the default provider
Configuration Parameters
| Setting | Description | Required |
|---|
| API Key | The provider's API key | Yes |
| Default model | The model used for speech synthesis | No |
| Default voice | The default voice ID to use | No |
| Default provider | Whether to set this as the default TTS provider | No |
How to Use
- Select the text content you want to convert
- Choose a voice
- The system sends the text to the TTS provider
- The generated audio can be played directly or saved
Voice Selection
Each TTS provider offers multiple voices to choose from. You can:
- View the list of available voices in the TTS settings
- Preview the effect of different voices
- Select one voice as the default
The voice list is fetched dynamically via API and will change as the provider updates its offerings.
| Format | Description |
|---|
mp3 | Universal audio format (default) |
wav | Lossless audio format |
ogg | Open-source compressed format |
opus | Efficient compressed format |
pcm | Raw audio data |
Synthesis Parameters
| Parameter | Description |
|---|
providerId | Specifies the TTS provider to use (optional) |
modelId | Specifies the model to use (optional) |
voiceId | Specifies the voice to use (optional) |
outputFormat | Output audio format (optional; defaults to mp3) |
outputDir | Save directory (optional) |
Frequently Asked Questions
| Question | Solution |
|---|
| ASR transcription is empty | Check that the audio file contains valid speech and is not corrupted |
| Unsupported audio format | Install ffmpeg to enable automatic format conversion |
| TTS-generated speech sounds unnatural | Try switching to a different voice or model; some voices perform better in specific languages |
| Voice list is empty | Confirm the API key is valid and the provider service is operating normally |
| Transcription result is inaccurate | Try specifying the correct language parameter, or select a more suitable model |
| TTS produces no audio output | Check the system volume and audio output device; confirm the generated file is not empty |