Recordings
Manage your organization's shared library of audio recordings — upload, transcribe, rename, and reuse them across every agent.
The Recordings page (/recordings) is your organization's shared library of audio clips. Recordings aren't tied to a single agent — upload one here and it's available to every agent in your organization.
This page covers managing the recordings library itself. For the full workflow of cloning a voice and building a hybrid agent that blends pre-recorded audio with LLM-generated speech, see Pre-recorded Audio.
Uploading a recording
Click Upload Recording to open the upload dialog. You can either:
- Choose audio files — any browser-supported audio format, up to 5 MB per file, up to 20 files per batch, or
- Record — record directly in your browser using your microphone. You'll be asked to name the recording first, then Start Recording; a timer and a Stop button appear while recording.
Talkr automatically transcribes each file as soon as it's added, and pre-fills an editable transcript field — review and correct it before uploading, since upload is blocked until every pending file has a transcript. You can also set a Language for the batch (defaults to multilingual auto-detect).
There's no duration limit on a recording — only the 5 MB file-size limit.
Managing recordings
Each recording in the list shows its ID (used for @mention references, see below), transcript, and original filename, with inline play/pause preview. Use the search box to filter by filename, transcript, or ID.
- Rename the ID — click the edit icon next to a recording's ID to give it a more memorable name (letters, numbers, hyphens, and underscores only). Talkr automatically updates every reference to the old ID across your agents' workflow definitions, so nothing breaks.
- Delete — removes the recording from the library after a confirmation prompt.
Using recordings in your agents
Recordings can be attached anywhere Talkr plays audio instead of synthesizing it with TTS:
| Where | How |
|---|---|
| Prompts | Type @ in any prompt field to insert a recording reference — see Pre-recorded Audio |
| Start Call greeting | Select a recording as the opening line for a Start Call node |
| Node transitions | Attach a recording as the transition_speech played when an edge fires |
| Tool custom message | An HTTP tool can play a recording (or speak custom text) before it makes the request — useful for a "let me check that for you, one moment" filler while the call is in flight |
| End Call goodbye | Play a recording as the closing line before the call ends |
| Transfer Call | Play a recording before transferring, or while Talkr resolves where to route the call |
Everywhere in the builder that lets you pick a recording uses the same searchable picker over this shared library, with inline playback so you can confirm you've selected the right clip.