| Method | How |
|---|---|
| Paste text | + Add → Import Transcript, paste into the text area |
| File upload | A local .txt, .vtt, .srt or .md, up to 10MB |
| Google Drive | Browse Drive once your Google account is connected |
01Three ways to get one in
02Tag the type
| Type | Use for |
|---|---|
| Interview | Customer discovery or research sessions |
| Sales call | Sales conversations |
| Support | Support calls or tickets |
| Other | Anything else |
The type tells Annsa how to weight and categorize what it finds.
03Add the customer, if you know it
Name shows up in Customer Voice quotes, email enables close-the-loop notifications when you ship, and revenue — what the customer pays you, as a number — scores the Revenue Growth lens (Annsa never invents this figure).
04Only the customer's words count
A labeled transcript can name who the customer is — tell Annsa the speaker and it extracts signals only from their turns. Your own team's talk track never becomes feedback, however quotable the pitch was.
05Formats it recognizes
Auto-detected, no configuration needed.
| Tool | Formats |
|---|---|
| Otter.ai | TXT |
| Fireflies | SRT, VTT |
| Grain | TXT |
| Whisper | TXT, SRT, VTT, TSV, JSON |
| Plain text | TXT |
06The limits
10MB per file, 50 to 500,000 characters when pasting, up to 25 segments extracted per transcript, and a 15-word minimum per segment.
07What happens next
Annsa splits the transcript into segments with speaker attribution where available, tags each source: transcript, and runs them through the same classification, grouping and scoring as everything else. Results land in Priorities alongside CSV, Slack and Survey feedback.
08It counts the same as any source
A quote from a call carries the same weight as a Survey submission.
09Uploading twice is safe
Transcripts are deduplicated on a content hash, so the same file uploaded again produces no duplicate feedback.