EditorLAB← Back to home

Artificial intelligence transparency

Last updated: October 2026

EditorLAB is an AI-assisted video creation service. The clips, subtitles, analyses and recommendations you get are produced or assisted by artificial intelligence systems. This page lists which ones, what they do, and what data they receive.

This is a translation of our French AI transparency page, which prevails in case of any discrepancy.

AI-generated content

The content produced by EditorLAB — clip cuts, subtitles, highlighted keywords and emojis, suggested titles and descriptions, analysis texts — is generated or assisted by AI. It may contain errors, approximations or omissions: a transcript can get a word wrong, a clip suggestion can miss the interesting moment, an analysis can be mistaken. Review and check before you publish.

If you share content made with EditorLAB, it is up to you to tell your audience that it was generated or modified by AI when the applicable rules require it.

The systems we use

Third-party providers are involved; part of the transcriptions is done on an EditorLAB server, with open models. The audio of your videos can therefore be transcribed on an EditorLAB server located in Belgium, or by AssemblyAI. EditorLAB does not train any model: data is sent to providers only for the time of processing.

Claude

by Anthropic

Analyzes the transcript text to suggest clips, write recommendations and answer questions.

Features involved
Clip suggestions · Help assistant · Translation
Data sent
The transcript text and the video metadata. No video or audio file is sent.
Models called
claude-sonnet-4-6, claude-sonnet-4-5, claude-haiku-4-5

AssemblyAI

by AssemblyAI

Turns speech into timestamped text, the basis of every analysis and of the subtitles, when transcription is not done on the EditorLAB server.

Features involved
Clip transcription
Data sent
The audio track of the video. For large files, only the extracted audio is sent, never the picture.
Models called
universal

Transcription EditorLAB

by EditorLAB (open models Whisper by OpenAI and Parakeet by NVIDIA)

Turns speech into timestamped text and separates the voices, like AssemblyAI, on an EditorLAB server located in Belgium. AssemblyAI takes over if it is unavailable.

Features involved
Clip transcription
Data sent
The audio track of the video, which stays on EditorLAB’s servers: it is not sent to any third party.
Models called
whisper-large-v3-turbo, parakeet-tdt-0.6b-v3, pyannote/speaker-diarization-community-1, speechbrain/spkrec-ecapa-voxceleb

What EditorLAB does not do

Your videos and transcripts are not used to train models on EditorLAB’s behalf. EditorLAB does not perform facial recognition, biometric identification, emotion recognition or scoring of the people filmed. Voice separation (“who speaks when”) tells apart the voices of a single video for subtitles and framing, without identifying anyone.

EditorLAB does not generate synthetic faces, voices or people, and therefore does not produce deepfakes. Clips are excerpts of your own files, reframed and subtitled.

Human decision

The AI suggests, you choose: selecting the excerpts, the settings and the export remain your own actions. Nothing is published without your consent: publishing to your social accounts, awaiting approval from TikTok, Google and Meta, only publishes the clips you schedule or, in automatic mode (Pro plan), the ones you have authorized.

Questions and complaints

For any question about the use of AI in EditorLAB, or to report a problematic result, write to service@editorlab.app.

See also the privacy policy and the terms of use.