From one recording to a ready-to-send delivery

Voiceover Delivery Cutter

Your edited WAV. Your client’s filenames. One checked delivery. Suggest cuts at quiet gaps, review every take, then package the right audio under the right names.

Local audio processing16 / 24-bit PCM WAVNo signup

1Bring your recording + delivery list

1. Edited PCM WAV

Mono or stereo · 16 / 24-bit · 8–192 kHz · up to 100 MiB and 10 minutes. Compressed and floating-point WAV are not supported.

No recording loaded

2. Ordered delivery CSV

First row: filename,prompt. Filename is required; prompt is optional. Up to 250 rows. Each included take is paired with the next row, in order.

Original synthesized test tones, not speech. Practice the workflow; the demo does not demonstrate spoken-prompt alignment.

Your audio, filenames and prompts stay in this browser. This page uses Google Analytics for canonical page visits only. No autosave; full offline operation is not promised.

Start with a WAV and your delivery list, or try the tone demo.

2Review the cuts

Silence detection only suggests boundaries. Listen around every start and end to protect quiet consonants, breaths and tails. No speech recognition or automatic prompt matching.

Load a WAV to inspect its waveform and edit takes.

Takes in delivery order

0

Select, reorder, split, merge or explicitly exclude retakes. Included takes pair sequentially with CSV rows. Reordering or excluding changes that pairing and clears your review confirmation.

No takes yet. The first import keeps the entire recording as one take.

    3Check every filename-to-take pairing

    0

    Matching counts do not prove matching speech. Read each prompt, listen to the assigned take, and check the names before confirming. Extra takes, missing takes or invalid/colliding filenames block export.

    Import a delivery CSV to see your checklist.

      4Build the delivery ZIP

      Adds zero-valued samples outside each retained source range, 0–2000 ms per end. This is separate from keeping existing source context at a suggested cut.

      Adds an apostrophe to formula-like text cells in the manifest CSV only. JSON keeps exact text; valid WAV filenames are unchanged.

      ZIP contains named WAV files, a CSV manifest and exact JSON manifest. Excluded takes are recorded. Exported WAVs preserve retained PCM samples, channels, sample rate and bit depth; original BWF, cue and other metadata chunks are omitted. Maximum output: 150 MiB.

      Add a WAV and delivery list to begin.

      A small tool for a specific handoff

      Made for e-learning lines, IVR prompts, course narration and other deliveries where a client has already supplied an ordered filename list. Start with an edited recording: this tool does not choose the best performance, remove noise or identify what was spoken.

      1. Import one edited PCM WAV and a CSV headed filename,prompt. Review the list before replacing any existing work.
      2. Optionally suggest cuts from quiet gaps. Inspect the staged ranges, then apply them explicitly. Or keep the full recording and split it manually.
      3. Listen to each boundary, adjust frame-quantized times and split or merge as needed. Exclude retakes explicitly and arrange included takes in the required order.
      4. Check each visible CSV-to-take pairing. Resolve every missing take, extra take and filename collision, then confirm your listening review.
      5. Choose any added digital silence, build the full ZIP and download it. Any subsequent edit invalidates the previous download. Keep your original recording.

      Before you deliver

      Can it recognize the prompt or find the best take?

      No. It measures audio levels, not meaning. Sequential pairing is a proposal based on your list order. You must listen and verify every filename, even when the counts match.

      What happens to quiet speech and long pauses?

      A quiet consonant or breath may fall below the threshold; a pause inside a sentence may look like a boundary. Suggested cuts never apply automatically. Use retained source context and manual boundaries, then audition both ends.

      Can I use MP3, stereo or a 24-bit WAV?

      Use little-endian RIFF PCM WAV, mono or stereo, 16- or 24-bit, at 8–192 kHz. Valid extensible PCM is supported. MP3, floating-point WAV, RF64 and compressed WAV are rejected. Up to 100 MiB, 10 minutes and 250 delivery rows.

      Are the exports bit-perfect copies?

      Retained PCM sample bytes are copied without normalization, fades, resampling or bit-depth conversion. Optional padding adds digital silence. Each WAV gets a new header and omits original BWF, cue and other metadata. Browser preview output may be resampled.

      What does the CSV need?

      Use a filename column and an optional prompt column, with comma, semicolon or tab separators. Names must be single safe filenames, never paths, and cannot collide, including case-only collisions. A .wav extension is added when absent. Quoted commas and multiline prompts are supported.

      Will my recording be uploaded or saved?

      No audio, filenames, prompts or tool actions are sent to analytics. Files are processed in browser memory with no autosave. Leaving, reloading or changing language can lose work. The page and local script assets still load over the network, and Google Analytics records a canonical page visit without query strings or referrer; offline availability is not guaranteed.