Recording Storage
Keep call recordings in your own S3-compatible bucket, under your retention policy.
Overview
Every call can produce three audio tracks — the caller, the agent, and a mixed recording — plus a transcript. By default Talkr stores these for you and deletes them after a short retention window.
If you need recordings kept longer, or kept under your own control, point Talkr at an S3-compatible bucket you own. Recordings are then written directly to your bucket as each call ends, and Talkr keeps no copy.
Recordings add up quickly. A team running 1,000 calls a day at three minutes each produces roughly 45 GB a month — about 540 GB over a year. Your own bucket keeps that under your retention rules, your compliance regime and your storage pricing.
Choosing where recordings are stored
| Talkr-managed (default) | Your own bucket | |
|---|---|---|
| Where recordings live | Talkr's storage | Your S3-compatible bucket |
| How long they are kept | A short retention window | However long your policy says |
| Who controls deletion | Talkr | You |
| Playback in Talkr | Yes | Yes |
| Extra cost | None | Your storage provider's |
The retention window that applies to Talkr-managed recordings is shown on the settings page itself, so you can see exactly what you are agreeing to.
Connecting your bucket
Go to Settings → Recording Storage and provide:
- Bucket — the bucket name, which must already exist
- Region — for example
us-east-1 - Endpoint URL — leave blank for Amazon S3; set it for MinIO, Wasabi, Cloudflare R2, Backblaze B2 or any other S3-compatible storage
- Access Key ID and Secret Access Key — credentials with permission to read, write and delete in that bucket
Use Test connection before saving. Talkr writes a small test object, signs a playback URL for it and deletes it again, then tells you exactly what worked. Saving runs the same check — a configuration Talkr cannot write to is never stored, because the alternative is recordings quietly going somewhere you did not intend.
Permissions your credentials need
s3:PutObject, s3:GetObject and s3:DeleteObject on the bucket's contents.
Read access matters as much as write: without it recordings upload correctly
but cannot be played back in Talkr.
Playing back recordings
Nothing changes in the Talkr interface. Open any call in Workflow Runs and the recordings play as usual.
Behind the scenes, Talkr generates a short-lived signed link and your browser fetches the audio directly from your bucket — the audio never passes through Talkr's servers.
Switching between storage options
You can switch at any time, and recordings already written stay where they are and remain playable. Talkr records where each individual recording was stored, so:
- Recordings written before you connected a bucket stay in Talkr's storage and follow its retention window
- Recordings written while your bucket is connected stay in your bucket, under your policy
- Switching back to Talkr-managed storage only changes where future calls are written
To go back, choose Use Talkr storage on the same settings page.
If your bucket becomes unreachable
Credentials get rotated and bucket policies change. If Talkr cannot write to your bucket when a call ends, it stores that recording in Talkr's storage instead rather than discarding it, and marks the call so you can see it happened.
Those fallback recordings follow Talkr's normal retention window — they are not kept indefinitely. Fix the credentials and subsequent calls resume writing to your bucket. Recordings that fell back are not moved retroactively.
Check Test connection after rotating keys. Talkr only discovers a broken configuration when a call tries to use it.
FTP and SFTP
Not supported today. S3-compatible storage is what allows playback to stream directly from your bucket; FTP has no equivalent of signed links, so every recording would have to be proxied through Talkr — which is what this feature exists to avoid. Most storage providers, including on-premise options like MinIO, offer an S3-compatible endpoint.