AI Transcription

The core feature of RiverScript is speech-to-text transcription. The service employs synchronized AI speech recognition models that operate in tandem to deliver higher accuracy and greater robustness when processing challenging audio segments.

RiverScript offers two synchronized speech recognition models: the Standard Sync Model (OpenAI Whisper v3 + Deepgram Nova-2) and the Premium Sync Model (ElevenLabs Scribe v2 + Deepgram Nova-3). Availability depends on the subscription plan. The features of the models are described in the Transcription models section.

To transcribe audio from the sources supported by RiverScript — audio and video files, in-app microphone recording, or Live Recording — users should navigate to one of the three main tabs in the main menu. At the top of the Files (Audio/Video), Microphone, and Live Recording windows are preliminary settings for the desired output format and input language. Selecting a specific input language typically reduces the likelihood of errors. If no input language is chosen, the AI model detects it automatically.

Before starting, it is recommended to select the preferred transcription type: Text Only or +Timecodes. In Text Only mode, the transcription is returned as structured text with paragraphs and punctuation. When +Timecodes is selected, the result includes clickable timestamps that can later be exported and used as subtitles, for example on YouTube.

The output settings also include two additional options — Auto-open and Speaker Diarization (available only for the Premium Sync Model). Auto-open is enabled by default. If a user needs to transcribe multiple files in succession and does not want the built-in editor to automatically switch between transcripts while they are still editing one, this option can be disabled. Speaker diarization is available exclusively in the Premium Sync Model. It separates different speakers voices within the transcript using labels that can be renamed directly in the editor.

Please note that AI-powered transcription is not guaranteed to be 100% accurate and may occasionally contain errors. We recommend reviewing all transcripts that require a high level of accuracy, regardless of the model used.