August update: Cloudflare R2, Hetzner, and MP4 to MP3

Two new places your files can land, and one new thing they can be. Here is what shipped.

Cloudflare R2 as a destination

You can now point any destination at a Cloudflare R2 bucket. R2’s headline feature is zero egress fees, which matters for exactly the workload a conversion API produces: files that get written once and downloaded many times. Converted output lands in your R2 bucket and serving it costs you storage, not bandwidth.

Setup is the same dashboard flow as every other provider, with one quirk worth knowing: R2 has no regions, so the region field holds your Cloudflare account ID instead. Converterer builds the endpoint as https://{account_id}.r2.cloudflarestorage.com from it. The credentials are an R2 S3 API token (access key ID and secret) with upload permission on the bucket, and there is no public/private ACL toggle because R2 rejects S3 ACLs; visibility follows your bucket’s own Cloudflare configuration.

Hetzner Object Storage, now documented

Hetzner Object Storage support actually shipped quietly earlier this year, but it never made it into the docs. Fixed. Three regions (fsn1 Falkenstein, nbg1 Nuremberg, hel1 Helsinki), a public/private ACL choice per destination, and the same S3-compatible credential pair as the rest of the family. If your infrastructure is already on Hetzner, your conversions no longer need to leave it.

That brings the destination roster to nine providers: AWS S3, Backblaze B2, Cloudflare R2, Hetzner, DigitalOcean Spaces, Linode, Vultr, Azure Blob, and Google Cloud Storage. Full credential details for each are in the destinations docs.

MP4 to MP3: the first audio output

The API can now extract the audio track from an MP4 and deliver it as an MP3:

curl -u "$CONVERTERER_API_KEY:" \
  https://api.converterer.com/convert \
  -F input=@webinar.mp4 \
  -F output_format=mp3

The usual jobs this covers: turning webinars and interviews into audio for transcription, repurposing video recordings as podcast episodes, and feeding speech-to-text pipelines that want audio, not video.

You get real control over the output: audio_qscale (0–9 VBR, 0 is best) or audio_bitrate for fixed-rate encoding, volume, channels (1 for mono is a common transcription trick: half the file size, no accuracy loss), sample_rate, and trim_start / trim_end for cutting intros and outros server-side. There is also audio_codec: copy, which remuxes the existing stream without re-encoding — but use it only if the source stream is already MP3. Most MP4s carry AAC audio, and copy will happily write that into a .mp3 file that many players reject. The default re-encode is the right choice unless you know otherwise.

To be precise about scope: this is MP4 (and MPEG) to MP3 only. Other video containers still convert to video, GIF, or still frames, not audio, and MP3 is the only audio output format. If your workload needs more routes than that, tell us which ones — this one exists because people asked.

The full option reference is in the conversion options docs, and the conversion page is at MP4 to MP3.