AudioWrench Audio Tools
Server Details
Reverse or trim selected WAV files, or transcribe WAV to TXT/SRT with Whisper Tiny and Base.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 23 tools
There are several parallel routes to the same outcome: aw.speech.transcribe (one-shot) overlaps heavily with the aw.speech.whisper_encode_step + aw.speech.whisper_decode_step manual pipeline, and aw.speech.transcript.export overlaps with aw.speech.whisper_export_txt (both yield TXT). aw.input.* lifecycle tools also collide with aw.jobs.cancel, which the input descriptions themselves have to point at. The extremely defensive, caveat-laden descriptions read as disambiguation-by-exhaustion rather than genuinely distinct tool purposes.
Names follow a stable dotted namespace scheme (aw.<domain>.<action>) with mostly verb-final segments: input.begin/commit/abort, jobs.list/cancel, session.open/close, transcript.read/export/clear. Minor deviations exist (aw.catalog, aw.status, aw.transfer.guide are noun-only, and audio.reverse/trim are verb-only), but the pattern is predictable enough to navigate.
23 tools is heavy for an audio-processing service; a meaningful share goes to two competing transcription pipelines rather than distinct capabilities. The session/input/artifact lifecycle justifies many of them, but the count sits at the borderline where consolidation is warranted.
The core lifecycle is covered: session open/close, staged input upload with verify/commit, audio trim/reverse, transcription, transcript read/export/clear, and artifact chunk/verify. But there is no enumeration of inputs or artifacts (read_chunk requires knowing what to ask for), no per-job detail beyond jobs.list, and no explicit way to discard a single committed input. Agents can work around this but will hit dead ends.
Available Tools
23 toolsaw.artifact.read_chunkARead-onlyInspect
Return one bounded binary embedded resource from this session artifact, plus range/whole-file hashes. Reassemble and hash ALL bytes. Not disk-save or whole-file verification from one range.
Inputs: offset (required), session (required), artifactId (required), length.
aw.artifact.verifyARead-onlyInspect
Re-read authenticated chunks and hash an actual current-session artifact; return metadata, never audio bytes.
Inputs: session (required), artifactId (required).
aw.audio.reverseBInspect
Reverse every channel with the actual AudioWrench core; verified new PCM WAV.
Inputs: source (required), session (required), operationId (required), float32.
aw.audio.trimCInspect
Copy the current source-frame interval using the unchanged core floor-to-frame semantics.
Inputs: source (required), session (required), endSeconds (required), operationId (required), +2 more.
aw.catalogCRead-onlyInspect
Find supported public beta workflows by task.
Inputs: limit, query, offset.
aw.input.abortADestructiveInspect
Discard this pending transfer attempt only; does not cancel the current processing job or delete a completed immutable input. Use jobs.cancel for all pending stages.
Inputs: attempt (required), inputId (required), session (required).
aw.input.beginAInspect
Reserve bounded encrypted staging for one explicitly selected file. NO source publication or job retirement until validated commit;2 pending/max24 identities.
Inputs: format (required), sha256 (required), inputId (required), session (required), +2 more.
aw.input.commitAInspect
Verify ALL staged bytes/hash/original format and atomically publish one current immutable input. Exact completed retries recover its receipt; invalid/partial data never retires a job.
Inputs: attempt (required), inputId (required), session (required).
aw.input.statusCRead-onlyInspect
Resume this exact transfer attempt without returning any selected bytes. Retired/complete state is explicit.
Inputs: attempt (required), inputId (required), session (required).
aw.input.write_chunkBInspect
Append one exact canonical-base6432768-byte chunk (last=remainder) to this attempt in order. Exact retries reconcile; binary bodies may reach client/model history.
Inputs: data (required), offset (required), attempt (required), inputId (required), +1 more.
aw.jobs.cancelBInspect
Retire active work AND in-flight uploads that have not committed. Completed immutable inputs/artifacts are not deleted. Stable upload UUID/operationId reconcile committed receipts. Bounded32 cancellation requests remain available after ordinary quota exhaustion; close clears active copies.
Inputs: session (required).
aw.jobs.listCRead-onlyInspect
Statuses for this session only, without selected audio or private URLs.
Inputs: session (required).
aw.session.closeADestructiveInspect
Explicitly revoke this transient remote session and remove its remote copies; original disk files are never touched.
Inputs: session (required).
aw.session.openAInspect
Create a private15-minute remote processing session. No local installation and no personal input is sent by opening it.
Inputs: no named parameters.
aw.speech.transcribeAInspect
Original Tiny/Base whole-file transcription with a CURRENT shared transcript. Explicit SAME-session original encoder handle required; Run at most64 original decoder steps for Tiny or32 for Base per explicit call, with identical source/model/language/output and the CURRENT handle returned by this same tool. Use a NEW operationId only when advancing; after a failed call reconcile jobs and the last accepted receipt before deciding whether to explicitly resubmit the exact same UUID and inputs. Never swap in the file-only decoder continuation. Pending is NOT text. On original completion atomically retain validated original segments in this session encrypted vault and return a whole TXT/SRT plus transcript handle. Read segments or export BOTH original formats without another inference; Clear is separate. Old whisper file tools unchanged. Same30s/TTL/storage/jobs/admission/quotas, no truncation, automatic heavy retry, device/OS recognition, E2EE/backup erasure/all139/capacity/accuracy claim.
Inputs: source (required), session (required), prepared (required), operationId (required), +4 more.
aw.speech.transcript.clearADestructiveInspect
Explicitly clear the EXPECTED CURRENT shared transcript only and leave an empty generation, matching original Clear semantics. Null succeeds only when no current state exists; a stale handle cannot clear a newer result. Does NOT cancel in-flight transcription, selected input, encoder/pending state, or previously downloaded immutable exports. A later valid model completion may populate the transcript again. Exact UUID replay must still refer to the same cleared generation. Removes active private encrypted snapshot, NOT irreversible provider backups/E2EE; no new model/device I/O.
Inputs: session (required), transcript (required), operationId (required).
aw.speech.transcript.exportAInspect
Export the CURRENT accepted transcript as exact original TXT or SRT, with NO model call. Same original serializers/timestamp fallback/CRLF, verified immutable header-auth artifact. Explicit current handle and UUID; replay requires same current generation/source/settings and actual verified bytes. Nothing to export after Clear/empty; stale/current-source mismatch rejects. A previously downloaded export is not erased by Clear. No browser OS-save/accuracy/backup-erasure claim.
Inputs: output (required), session (required), transcript (required), operationId (required).
aw.speech.transcript.readARead-onlyInspect
Read CURRENT accepted original transcript segments and exact time values, explicitly, in bounded pages. Null transcript selects current state; an expected handle fences stale/cleared/newer results. No model I/O, recording or unfinished text. Remote selected transcript text is returned only by this explicit read. Empty state is useful; no editing or calibrated alignment claim. Same session/source epoch/engine/expiry/privacy bounds.
Inputs: session (required), limit, offset, transcript.
aw.speech.whisper_decode_stepBInspect
Explicit bounded continuation of the REAL original Tiny/Base decoder for a whole accepted selected WAV. First call uses the SAME-session original encoder handle and null continuation. A nonterminal step returns only an opaque encrypted SAME-session decoder-pending handle, NOT a transcript; call this tool explicitly again with identical source/prepared/model/language/output, the CURRENT returned handle and a new operationId. Original complete token history, logits/stopping rules and full TXT/SRT remain unchanged; each call yields after at most64 original nonterminal steps for Tiny or32 for Base and keeps the existing30s work limit, session caps, TTL/admission and quotas. Rebuilds constant encoder KV with one original initial decoder forward and verifies retained exact hashes. Base reprojects six original values from the same pinned encoder source as needed; no new encoder run. Atomically replaces old private state; cancel/failure preserves the last accepted state. Consumed pending handles fail SUPERSEDED, not automatic heavy retry. No cache upload/download/inline tensors, model change, truncation, native CPU preemption, accuracy/all139/physical capacity/client disk access or provider acceptance claim. Auto retains original English default; original padded timestamps require review.
Inputs: source (required), session (required), prepared (required), operationId (required), +4 more.
aw.speech.whisper_encode_stepAInspect
Prepare the exact original Tiny or Base encoder in separate bounded calls: two for Tiny, three for Base. Begin with the accepted WAV source and no continuation. While status is encoder-pending, supply its opaque continuation with the same session, source and model and a new operationId. Retry the same operationId with identical arguments only; no automatic model retry. After encoder-prepared, choose decoderRoutes.fileOnly for a whole file without replacing the current transcript, or decoderRoutes.currentTranscript for a whole file plus shared current segments, export and Clear. Both use the same prepared handle, session, source and model; start with null decoder continuation. Never switch tools with a pending decoder continuation. Intermediate activations stay in the encrypted session; no upload, download or transcript is produced by this tool.
Inputs: source (required), session (required), operationId (required), model, +1 more.
aw.speech.whisper_export_txtAInspect
Export the original TXT from a complete current-session SRT produced by aw.speech.whisper_decode_step, without invoking the model again. Only a verified original decoder SRT artifact in the same active source revision is accepted; no imported SRT, other session, stale source, or unfinished state. Keeps the original whole-second timestamps, text and CRLF. Returns a verified immutable authenticated TXT artifact, not inline transcript content. Reuse operationId for the same SRT; inspect the job before an explicit retry.
Inputs: srt (required), session (required), operationId (required).
aw.transfer.guideCRead-onlyInspect
Explicit native MCP file-transfer contract, bounds and client-history privacy. Discovery does not read your disk.
Inputs: no named parameters.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
49 audio tools: trim, convert, normalize, remove vocals, split stems, transcribe, and more.
Cut audio from YouTube, TikTok or Instagram links to MP3/WAV, and fetch video transcripts.
Transcribe audio and video files to text with timestamps and SRT or WebVTT subtitles
Whisper speech to text, 99 languages: files, TikTok and podcast links to text, SRT, Word/PDF.
Related MCP Servers
- AlicenseAqualityBmaintenanceAudio perception for AI agents. Measure loudness, clipping, noise and pitch; view spectrograms; split audio into events; transcribe speech; compare before/after; and clean up, edit or synthesize sound. Runs locally, no API keys.14119 PyPIMIT
- AlicenseNot gradedqualityDmaintenanceExposes audio processing tools (inspect, normalize, trim, fade, gain, filter, speed, reverse) as MCP tools, enabling Claude to manipulate WAV files through natural language commands.MIT
- FlicenseNot gradedqualityCmaintenanceEnables high-performance audio transcription using Faster Whisper with CUDA acceleration, supporting single and batch audio file processing with multiple output formats (VTT, SRT, JSON).-
- AlicenseNot gradedqualityCmaintenanceEnables local transcription of audio/video files and YouTube URLs, generation of SRT/VTT subtitles, and analysis of speech pacing and audience retention risk using faster-whisper.MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.