Hebrew & English transcription. Optimized for Hebrew.

This is an AI-agent skill designed to work with OpenAI Codex, Claude Code, and other compatible AI-agent environments. Transcription runs in the cloud — if it's a local file, the skill uploads it to our servers and returns the transcript. If it's a Drive, S3, or other cloud link — it fetches the file directly. If it's a YouTube, Facebook, Instagram, or Twitter/X link — it downloads it automatically. Optimized for Hebrew, with full English support.

Any publicly accessible link — yes. Google Drive, S3, and more. Links that require authentication (internal sharing) are not supported.

Yes. On Basic and Premium plans you can upload files up to 2 GB. On the Enterprise plan, up to 5 GB.

TextOps is designed to work with Codex, Claude Code, and other compatible AI-agent environments (any environment that supports the SKILL.md format and can run Python). Uploading local files and fetching from cloud sources works the same way in all of them. For YouTube, Facebook, Instagram, and Twitter/X — since some platforms block certain IP ranges, there may be limitations when the agent itself runs in the cloud rather than on your local machine.

Yes. TextOps can be used from Codex to transcribe local audio and video files and supported sources, then continue in the same workflow with summaries, documentation, subtitles, show notes, action-item extraction, or further agent processing. Support for some URL sources can depend on the environment in which the agent runs and its network restrictions.

Audio and video files are automatically deleted after one day. On the Enterprise plan, a custom retention policy can be arranged.

The model automatically detects the number of speakers. However, you can always specify that it's a single speaker — it will skip the detection step, which is much faster. For summaries and supplementary content, transcribing as a single speaker is usually sufficient.

Hebrew is the default. If the recording is in English, just mention it — for example: "this is English" — and the model will transcribe accordingly.

The service supports Hebrew and English. Transcription is optimized for Hebrew, with a focus on high accuracy and spoken Hebrew.

Absolutely! The transcript itself doesn't enter the context unless explicitly requested. The model only reads important messages to save tokens. The skill is designed so you can continue using the model while it runs in the background. There are minor differences across IDEs and environments, but the skill is built to support harness development.

Ask the skill itself! It's guided to explain what it does and how to use it effectively. In general, the more detail you provide — whether it's a single speaker, whether you want timestamps, or if it's for subtitles — the faster and more efficient the process will be.