Convert Speech into Text via Google's Gemini API
Transcribe recorded speech into text within a CNAPS Studio flow. Transcribe (Gemini) was added in v1.2.2. Requests run through Google; your configured Google/Gemini API key authorizes the request and is billed to your provider account.
🔑 Requires a Google / Gemini API key
What it does
Transcribe (Gemini) converts recorded speech into text within a CNAPS Studio flow. It is an external model introduced in the v1.2.2 release. Use it to turn spoken material into a readable transcript, then connect that text to a language model for a separate summary or rewrite. Requests run through Google rather than on CNAPS-hosted GPUs.
Problem it solves
- Readable transcripts — Turn recorded speech into text you can review and search.
- Notes from spoken material — Feed a transcript into a separate summarization stage.
- Reusable transcription flows — Apply the same processing steps to repeated recordings.
- External transcription option — Use your existing Google integration alongside other Studio nodes.
Input/Output
- Input: Speech audio supplied to the transcription node. Extract the audio first when starting from a video.
- Output: Transcription text for Text Viewer or downstream text processing.
- Requires: A Google/Gemini API key configured in Workspace Settings → Integrations.
Configuring the Node
Speech input — Use a clear recording with as little competing music or noise as practical.
Language and transcription options — Use the options exposed by the selected node configuration.
Timing and speaker information — Check the output format before connecting it to subtitle tools; plain text is not a word-timed transcript.
How to Use
- Configure your Google/Gemini API key under Workspace Settings → Integrations.
- Upload the recording with Sound Loader. For a video, use Video Audio Extract first.
- Connect the audio to Transcribe (Gemini) and the text output to Text Viewer.
- Run and review names, numbers, technical terms, and speaker changes.
- Connect the reviewed text to a language model if you need a summary, translation, or structured notes.
Example Workflow
Flow: Sound Loader → Transcribe (Gemini) → Text Viewer.
Video source: Video Loader → Video Audio Extract → Transcribe (Gemini) → Text Viewer.
For notes, add a language model after transcription with the instruction: “Summarize the transcript in five bullet points and list any action items.”
Accuracy & Speed
Execution | Google API; no local GPU required. |
Billing | Provider charges apply to your Google account; CNAPS usage capacity is not consumed. |
Accuracy | Noise, overlapping speakers, accents, and specialist vocabulary can affect the transcript. |
Output compatibility | Confirm the actual response format before using timestamps or speaker data downstream. |
How it compares
Speech Recognition (Clips) and Speech Recognition (Video) provide other transcription paths in Studio. Choose the node by input modality, required output structure, and provider requirements. Transcription, summarization, and subtitle rendering are separate stages.
Technical Details (for engineers)
Google documents Gemini audio understanding and a dedicated Gemini Transcribe model. Upstream capabilities include structured transcription features, but the exact fields available in a Studio flow depend on its node implementation.
Model Source
- Gemini — official audio guide: Gemini — official audio guide
- Gemini Transcribe — official model page: Gemini Transcribe — official model page
- CNAPS integration: v1.2.2 release note
Compliance & Provenance
Provider | Google |
License / service terms | Google API terms and the terms applicable to your provider account. |
Output provenance | A transcript is extracted speech content; see CNAPS AI Policy for applicable attribution and metadata. |
For platform classification, acceptable use, and output marking, see the CNAPS AI Policy.
Run it now
Open CNAPS Studio, find Transcribe (Gemini) in the Component panel, and start with the example workflow above.