A launch video filmed at a Chinese supplier may move through a local production crew, an overseas editor and several regional marketing teams. “Recorded in 32-bit float” tells those teams something about a file. It does not settle whether the recording is usable or whether the next team can handle it correctly.
For procurement, ONCE proposes three acceptance stages: the original recording, an adjusted working copy and the delivery export. Use them to assign responsibility before treating an audio format as a staffing saving.
Separate file headroom from distortion earlier in the chain
Sound Devices explains in its Understanding 32-bit Float, updated January 19, 2026, that recorded values above 0 dBFS can be attenuated in a compatible audio workstation. That describes recoverable file levels, rather than a guarantee about every upstream component.
DPA’s 32-bit floating-point and other conversions identifies a separate limit: an analog input can still overload if the microphone output exceeds what it accepts. Floating-point storage cannot reconstruct signal already lost there.
Monitoring is another distinction. In its NT1 5th Gen clipping guide, updated July 22, 2026, RØDE distinguishes monitoring clipping from recorded distortion, including distortion beyond the microphone’s sound-pressure capability. Its instructions concern that model; they are not universal settings for other recording systems.
Stage one: retain a sample from the production setup
The following is a proposed workflow, not a report of an ONCE client test. Record a short rehearsal using the intended microphone, connections, placement, clothing and recording settings. Include quiet speech, ordinary delivery and the loudest intended performance. Capture product actions under normal planned operating conditions; there is no need to manufacture extreme sounds.
Imagine a supplier demonstration with a spoken explanation and a mechanical closing action. Both belong in the rehearsal. If the crew later changes microphone position or adds a wireless link, log the change and decide whether the sample needs repeating.
Ask the sound lead to identify each track, its signal path, the recording format and the person who checked the original. A floating-point file header alone does not establish what happened before that file was written.
Stage two: assess an adjusted copy
Keep the original untouched. Open a working copy in a workstation that supports the format, bypass processing that could obscure the check and reduce file or clip gain as needed. Listen at a comfortable playback level. Lowering headphone volume alone is not the same operation.
| Check | Question | Decision |
|---|---|---|
| File level | Does suitable clip attenuation make the passage usable? | Retain the adjustment record; do not reject solely on an over-range display |
| Processing and playback | Is the original usable but a processed or monitored path problematic? | Investigate that path and compare again |
| Capture chain | Does distortion remain in the adjusted copy? | Have the sound lead investigate and rehearse again where possible |
| Editorial use | Are there clothing noises, interruptions, missing words or mispronunciations? | Agree a pickup or edit; the format is not a pass criterion |
This is a triage sequence, not a way to identify a faulty component by listening alone. Persistent distortion needs investigation. Report the timecode, settings and changes rather than immediately blaming the microphone.
Stage three: reopen what the client will receive
Export in the agreed delivery format and check that actual file, as well as the finished video. Log format, sample rate, channels, level processing and version. Set requirements for the intended use; a recording-format label cannot replace an export specification.
Keep source recordings distinct from working projects and final deliverables. DPA warns against overwriting originals when converting to fixed-point files, because the floating-point representation is no longer retained. The handover should identify both the source track and the version actually used.
For a brand commissioning production through a China-based partner, agree who receives source audio, how long it is retained and who can access it. If another regional editor needs dialogue or product sounds separately, include those stems in the scope before the shoot. Further production planning can start with the ONCE video production articles.
Compare responsibility before comparing format labels
For an event that is difficult to repeat, prioritize the monitoring, recording checks, contingency arrangements and accountable sound lead. For repeatable studio narration, schedule a rehearsal and a post-production compatibility check. Neither case justifies automatically removing sound work because of a format change.
Ask suppliers to separate preparation, capture, post-production and source delivery in their scope. If a proposal removes work, identify the exact step, its acceptance evidence and who handles a failure. No universal savings percentage follows from bit depth.
A useful procurement sentence is: “Rehearse the actual recording path and provide the untouched source, an approved adjusted version and the agreed export, with unresolved issues identified.” It gives the receiving editor something concrete to verify.
Source scope
Official sources were checked on October 28, 2026. DPA’s article, by Eddy Bøgh Brixen, shows no publication date. These sources establish technical boundaries; they do not demonstrate staffing reductions, production savings or results from an ONCE project.
