Import a transcript — how to use it in the Studio
Bring the transcript from another system (text, subtitles or the JSON of several services) and use it in the FalaAI Studio without transcribing again. The format is recognised from the file itself and everything becomes the same dialogue, with speakers and timestamps when they exist.
The best path
You do not need to transcribe twice — pick the one that matches what you already have:
- Already have the transcript (Whisper, Teams, Zoom, PBX, another system)? Import it here: it is free and the file is not sent to any server.
- Only have the audio? Use the Audio card: the transcript comes ready, with speakers and audio events.
- Have a WhatsApp chat export? Use the Chat extraction card.
Step by step in the Studio
- In the FalaAI Studio, Input tab, My text / other system card: drop the file on the dashed area or click it to choose.
- The format is recognised from the file itself (text, subtitles or JSON). The Service selector is there to load a ready example.
- Check the dialogue and the speakers that appeared. If the file does not identify who spoke, the text comes in as a single block.
- The content stays loaded in the Input and goes on to the Diagnosis and the Audit (Load from Input).
- The character count appears in the panel — it is what defines the cost of the analyses.
Accepted formats
Plain text, subtitles with timecodes and the response JSON of seven transcription services (besides ours). Click each one to see how it works and open an example file.
Plain text (.txt)
A running transcript. If lines start with Speaker N: (or Interlocutor/Hablante), they become a dialogue with speakers; otherwise it is treated as a single text.
SRT (.srt)
SubRip captions: index, a time line HH:MM:SS,mmm --> HH:MM:SS,mmm and the text (optionally with Speaker N:).
WebVTT (.vtt)
WebVTT captions: WEBVTT header and cues with HH:MM:SS.mmm timing.
AssemblyAI
Uses utterances[] (speaker/time/text) and words[]. Times in milliseconds.
AWS Transcribe
Uses results.transcripts[], results.items[] and results.speaker_labels.segments[]. Times as string seconds.
ElevenLabs Scribe
Uses words[] with type (word/spacing/audio_event) and speaker_id. The only commercial one that carries audio events.
Deepgram
Uses results.channels[].alternatives[] and results.utterances[] (diarization by speaker 0/1).
Google Cloud STT
Uses results[].alternatives[].words[] with startTime/endTime as string seconds (e.g. 3.500s) and speakerTag.
Azure Speech
Fast transcription: phrases[] with speaker, offsetMilliseconds and durationMilliseconds.
OpenAI Whisper
verbose_json format: segments[] with start/end/text. No native diarization (single speaker).
What you need to know
What it costs
- Importing is FREE: the file is read in your browser and nothing is sent.
- The charge happens later, when you generate the Diagnosis or the Audit — charged per character.
- The Studio shows the character count of what you imported (valid content only, without labels, times or events).
- See the rates and rules in the Diagnosis and Audit help pages.
What is not accepted
- Empty file.
- Invalid JSON or corrupted file.
- Unrecognised format (not text, not subtitles and not the JSON of a supported service).
- Audio file or WhatsApp chat export: those go in the Audio and Chat extraction cards.
The example files are on this page, in each format (Open example) — use them to test before importing yours.