Transcription Automation
Recordings become clean, labelled transcripts.
Transcription Automation
Voice & Media
- RuleDone:
A recording arrives from Drive, Zoom or a portal upload.
- RuleDone:
Consent, file type, size, sensitivity and the daily hour budget are checked.
- RuleDone:
The audio is prepared safely and sent to the speech provider allowed for its sensitivity.
- RuleDone:
Low-confidence words, crosstalk and gaps are marked, and temporary copies are deleted.
- RuleDone:
An optional summary is written and checked against the transcript.
- RuleWaiting for approval:Waiting for you
After any needed review, the files are saved to your folder.
For example: A training company records a 90-minute workshop in Zoom.
Waiting for you. Nothing has changed yet.
Review for important actions. Customer-facing and record-changing actions are reviewed first.
What is Transcription Automation?
In shortTranscription Automation is an AI automation for Voice & Media teams that works with AssemblyAI, Deepgram, Google Drive, Indian-language speech providers, Whisper (self-hosted) and Zoom. Turns uploaded audio and video into timestamped, speaker-separated transcripts and optional summaries in your folder, and deletes the temporary copies. It is built for operations, research, training and media teams and legal-support staff.
Last reviewed
Key facts
- Department
- Voice & Media
- Starts when
- An audio or video file lands in a watched Drive folder, a Zoom cloud recording finishes, or someone uploads one in the portal
- Works with
- AssemblyAI, Deepgram, Google Drive, Indian-language speech providers, Whisper (self-hosted), Zoom
- Run modes offered
- Shadow, Draft, Assisted and Auto
- Human control
- Review for important actions
- Status
- Available now
The problem today
Staff pay per-file services by hand, download and upload files, rename them, and forget to delete sensitive recordings. Indian-language and mixed-language audio is handled poorly.
What it does
Every consented audio or video file becomes a timestamped transcript with speakers separated as Speaker 1, Speaker 2 and so on. You can add an optional summary with chapters. Transcripts are saved to your chosen folder as Word, text, subtitle and JSON files, each clearly marked "machine-generated". The temporary audio and the provider's copy are deleted as soon as the transcript is stored.
How does Transcription Automation work?
How it works- Rule you set
- AI step
- A person decides
- Step 1Rule
A recording arrives from Drive, Zoom or a portal upload.
- Step 2Rule
Consent, file type, size, sensitivity and the daily hour budget are checked.
- Step 3Rule
The audio is prepared safely and sent to the speech provider allowed for its sensitivity.
- Step 4Rule
Low-confidence words, crosstalk and gaps are marked, and temporary copies are deleted.
- Step 5Rule
An optional summary is written and checked against the transcript.
- Step 6Rule
After any needed review, the files are saved to your folder.
What you get
Key features- Outputs: .docx, .txt, .vtt, .srt and .json
- Speaker separation, with real names only when the uploader types them in
- Unclear words marked [?word], plus [crosstalk] and [inaudible] tags
- Custom vocabulary from your glossary
- Optional summary with topics and timestamped chapters
- Daily audio-hour cap to control cost (default 50 hours)
How much does it do on its own?
Run modesPick a mode to see what Transcription Automation would do at each step.
Every automation starts in Shadow. Move it up when you are ready, and back down at any time. How run modes work
Transcription Automation in Shadow
Runs on real data and shows what it would have done. No app is changed and no message is sent.
- Rule
A recording arrives from Drive, Zoom or a portal upload.
Watches for it
- Rule
Consent, file type, size, sensitivity and the daily hour budget are checked.
Runs on real data, read only
- Rule
The audio is prepared safely and sent to the speech provider allowed for its sensitivity.
Runs on real data, read only
- Rule
Low-confidence words, crosstalk and gaps are marked, and temporary copies are deleted.
Runs on real data, read only
- Rule
An optional summary is written and checked against the transcript.
Runs on real data, read only
- Rule
After any needed review, the files are saved to your folder.
Shown to you as a suggestion
Shadow vs Auto for Transcription Automation
Auto is the most it can do on its own. Here is each step side by side.
| Step | Shadow | Auto |
|---|---|---|
| In short | Runs on real data and shows what it would have done. No app is changed and no message is sent. | Runs without approval, with limits, alerts and a kill switch. Unlocks after a trial period in Assisted. |
| Step 1A recording arrives from Drive, Zoom or a portal upload. | Watches for it | Runs on its own within your limits |
| Step 2Consent, file type, size, sensitivity and the daily hour budget are checked. | Runs on real data, read only | Runs on its own within your limits |
| Step 3The audio is prepared safely and sent to the speech provider allowed for its sensitivity. | Runs on real data, read only | Runs on its own within your limits |
| Step 4Low-confidence words, crosstalk and gaps are marked, and temporary copies are deleted. | Runs on real data, read only | Runs on its own within your limits |
| Step 5An optional summary is written and checked against the transcript. | Runs on real data, read only | Runs on its own within your limits |
| Step 6After any needed review, the files are saved to your folder. | Shown to you as a suggestion | Runs on its own within your limits |
| True in every mode | No file is transcribed without a consent attestation for its source. | |
What it will never do on its own
Safety & human control- No file is transcribed without a consent attestation for its source.
- Never identifies people by their voice, and never guesses speaker names.
- Health audio is refused until a BAA is signed. Confidential files use self-hosted Whisper by default.
- Confidential, health and low-confidence files always need the uploader's or folder owner's approval.
- Deleting the original is off by default and needs two different admins to approve.
- You approve before it acts
- Your data stays yours
- It knows its limits
- Stop anytime
What will you see on your dashboard?
Results you can trackTargets are quality goals the automation is built to reach. Estimates are calculated from your own volumes.
- Target
Hours transcribed and turnaround (target: p95 within 0.5× audio length + 5 minutes)
A goal it is built to reach
- Target
Word error rate (target: ≤ 12% English, ≤ 20% Indian languages)
A goal it is built to reach
- Tracked
Cost per audio hour, and share needing human review
Measured on your own data
- Tracked
Average word confidence
Measured on your own data
- Target
Temporary files purged on time (target: 100%)
A goal it is built to reach
A training company records a 90-minute workshop in Zoom. When the recording finishes, a transcript with four speakers and a 12-chapter summary lands in the course folder within about 45 minutes. The trainer types the speaker names on the review screen before it is saved.
Who is Transcription Automation for?
- Operations
- Research
- Training and media teams
- Legal-support staff
Industries: Professional services, Education & training, Agencies, Healthcare & clinics, Legal firms, Nonprofits
Also useful for: Productivity & Projects
Connects to the apps you already use
Works withApp names are trademarks of their owners. Charviam connects to them; it is not a partner or reseller.
- AssemblyAI
- Deepgram
- Google Drive
- Indian-language speech providers
- Whisper (self-hosted)
- Zoom
Better together
Automations that hand work to Transcription Automation, or pick up where it stops.
What do you need to set it up?
Setup- Connect Google Drive (pick folders), Zoom and a speech provider
- Confirm recording consent for each source
- Choose a destination folder per source, output formats and languages
- Tag sensitivity per folder (normal, confidential, health)
- Run Shadow for 2 weeks and correct 10 hours of audio per language
Typical setup
- 01Connect your apps
- 02Configure the rules
- 03Run a watch-only trial
- 04Go live
Not included
- Live captioning or joining meetings
- Legal-grade or court transcripts
- Translation (use 40) or identifying speakers by voice
Questions about Transcription Automation
FAQSee Transcription Automation on your own data
WalkthroughIt starts in Shadow, where it only watches and suggests, so nothing changes in your apps until you say so.