Public declaration syntax from forge-rs/crates/forge-media/src/transcription.rs
Original source SHA-256: 0d553e0ce09c44ab7d4599d73d30579be9edb08e6f71a561b51d978df8a484a3
Function bodies and constant values are omitted. This is not the complete implementation.

Source line 54
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
pub struct TranscriptionSegment {
/// The start time of this segment in seconds from the beginning of the audio.

pub start: f64,
/// The end time of this segment in seconds from the beginning of the audio.

pub end: f64,
/// The transcribed text for this segment.

pub text: String
}

Source line 78
pub fn duration(&self) -> f64;

Source line 101
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
pub struct TranscriptionOptions {
/// An ISO 639-1 language code hint (e.g., "en", "fr", "de").

/// `None` means the provider will auto-detect the language.

pub language: Option<String>,
/// An optional prompt to guide the transcription model. Useful for providing

/// context about domain-specific terminology or expected content.

/// `None` means no guidance prompt.

pub prompt: Option<String>
}

Source line 142
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct TranscriptionResult {
/// The full transcribed text.

pub text: String,
/// The detected language as an ISO 639-1 code, if available.

pub language: Option<String>,
/// The total duration of the audio in seconds, if available.

pub duration_seconds: Option<f64>,
/// Time-aligned transcription segments. May be empty if the provider

/// does not support segmented output.

pub segments: Vec<TranscriptionSegment>
}

Source line 198
#[async_trait]
pub trait TranscriptionProvider: Send + Sync {
    /// Returns the model identifier (e.g., "whisper-1", "chirp-v2").
    fn model_id(&self) -> &str;

    /// Returns the provider name (e.g., "openai", "google").
    fn provider_name(&self) -> &str;

    /// Transcribes audio data to text.
    ///
    /// # Arguments
    ///
    /// * `audio` - Raw audio bytes. The provider determines acceptable formats
    ///   (e.g., WAV, MP3, FLAC, OGG).
    /// * `options` - Configuration for language hint and guidance prompt.
    ///
    /// # Returns
    ///
    /// A [`TranscriptionResult`] containing the transcribed text and metadata.
    ///
    /// # Errors
    ///
    /// * [`ForgeMediaError::TranscriptionFailed`] -- if the provider returns an error.
    /// * [`ForgeMediaError::UnsupportedFormat`] -- if the audio format is not supported.
    async fn transcribe(
        &self,
        audio: &[u8],
        options: &TranscriptionOptions,
    ) -> Result<TranscriptionResult, ForgeMediaError>;
}

Source line 273
pub async fn transcribe(
    provider: &dyn TranscriptionProvider,
    audio: &[u8],
    options: &TranscriptionOptions,
) -> Result<TranscriptionResult, ForgeMediaError>;