New
Video and Audio Understanding
- Upload video and audio — Attach a video (MP4, MOV, M4V or WEBM, up to 10 minutes) or an audio file (MP3, M4A, MP4, WAV, OGG, OPUS, FLAC, AAC or WEBM, up to 2 hours) of up to 100 MB in the chat and ask about it. See examples in Video Understanding AI.
- Sees and hears — Risphai looks at a video one frame per second and transcribes its audio with timestamps, so answers can say what happens, what is said and when.
- Any model — Every model can use the transcript, and the frames go to a model that can analyze images.
- Every plan — Video and audio uploads work on every plan, in regular and incognito chats.
- Video previews — The first frame of a video is shown as its preview in the chat box and in your sent message.
- Faster uploads — Transcription starts as soon as the audio is ready, and the frames and transcript are reviewed while the rest of the video is still being prepared.
Content Screening
- Checked before upload — The frames, the preview image and the transcript are checked by OpenAI's moderation service, and the video file is uploaded only after everything passes.
- Rejected uploads — If content is flagged or the check cannot be completed, the upload is rejected. An audio file whose transcript is flagged is deleted immediately.
Pricing
- Transcription — 50 credits per hour of audio, calculated from the length of the file, so a 10-minute video costs about 8.34 credits. Videos without sound are not transcribed and have no transcription cost.
- Refunds — Credits are returned if transcription fails. They stay charged if the finished upload is then rejected by content screening.
File Manager
- New sections — Uploaded videos and audio files now have their own Uploaded Videos and Uploaded Audios sections in the File Manager, with previews and deletion.
Changes
Transcription
- Tip in the transcription window — The transcription window now points out that you can attach an audio file directly in the chat box to get a timestamped transcript for the AI.
- More reliable — Transcription now retries automatically when the connection to the speech-to-text provider drops.
Privacy and Legal
- Updated documents — The Privacy Policy, Terms of Service and Data Protection Summary now describe video and audio uploads, transcription by xAI, content screening by OpenAI, the cleanup of unfinished uploads after 24 hours and the rules for uploading recordings of other people. The Terms of Service now include the transcription price and are version 2026-10-17.
Current models and prices are on the Information page.