WonderPhi Phone -- Telephony & Voice Batch Recipes
Publicly online since 2010 · U.S. patent applications since 2012 · inventions offered since 2014. The work of Christopher Gabriel Brown, independently documented.
WonderPhi Phone -- Telephony & Voice Batch Recipes
28 recipes for voicemail transcription, call recording processing, CDR parsing, phone E.164 normalization, DNC scrub, sudo-SMS queue writing. No Twilio. Requires WonderPhi Compute.
© CRI-ONE. All rights reserved. Patents issued and pending. Unauthorized reproduction of the underlying designs is prohibited.
Extended catalog & full narrative — WonderPhi Phone -- Telephony & Voice Batch Recipes
The extended dossier appended from the 2026-08-05 catalog snapshot. Prices in the body copy have been stripped; the live-store price on this page is the authoritative figure. Images have been omitted.
Easy Setup — No Package Manager, No Account, No Cloud
Single Python file. Python 3.8+ standard library only — nothing to pip install. Run with:
python wpp-recipe.py
Sample plan files ship with the download and run out of the box. Pair with WonderPhi Compute to execute the generated plans in parallel.
WonderPhi Phone — 28 batch recipes for telephony & voice operations
Twenty-eight recipes for voicemail-heavy, call-center-adjacent, and audio-heavy workflows: Whisper transcription of hundreds of voicemails at once, CDR (call-detail-record) parsing, E.164 phone-number normalization, DNC (Do-Not-Call) list scrubbing, audio format conversion, sudo-SMS queueing, hold-music preparation.
Why this exists
Anyone who runs a phone system ends up with folders of audio files, CDR exports from carriers, and lists of phone numbers that need cleaning. Transcribing 400 voicemails by hand is nobody's job. Scrubbing 20,000 phone numbers against the DNC list before you dial them is legally required and manually miserable.
These recipes wrap the tools that do each of those jobs (Whisper, ffmpeg, standard number-parsing libraries reimplemented in pure Python) so you can point at a folder and walk away.
What it is, in plain English
Each recipe knows one voice-adjacent job. Transcribe every voicemail in a folder. Convert every audio recording to mp3. Parse today's CDR export into a per-caller summary. Normalize a mixed-format phone list to E.164. Compare a call list against the DNC registry. The recipes emit plan files; Compute fans out across cores.
How you use it — three steps
- Run the recipe file.
python wpp-recipe.pyin the folder that contains your input files, or wherever you want the plan written. - Point at the target. The recipe asks (or reads from a text file) what the input is — a folder, a list of hostnames, a CSV, whatever the recipe expects. It writes a plan file: one shell command per line.
- Hand the plan to Compute. Drop the generated
plan.wpc.txtinto the WonderPhi Compute inbox, and every command runs in parallel across your cores. Read the honest log when it's done.
Worked examples — three of the 28 recipes in this library
whisper-transcribe-folder — transcribe every voicemail
One .txt transcript per audio file. Runs OpenAI's Whisper (or whisper.cpp) locally — no cloud upload.
python wpp-recipe.py whisper-transcribe --input ./voicemails --output ./transcripts --model small.en
cdr-parse-carrier — turn a carrier CDR export into a per-caller summary
Reads the CDR CSV (common formats: Twilio, Bandwidth, Cisco), aggregates by caller, emits a summary CSV.
python wpp-recipe.py cdr-parse --input cdr-2026-07.csv --format twilio --output cdr-summary-2026-07.csv
e164-normalize — clean up a mixed-format phone list
Turns (770) 776-7023, 770.776.7023, +17707767023, 7707767023 all into +17707767023.
python wpp-recipe.py e164-normalize --input contacts-raw.csv --output contacts-clean.csv
Verified capabilities
Recipes work on any Windows / macOS / Linux workstation. Whisper recipes need a machine with 4+ GB RAM and either CPU or GPU:
- 28 operations covered: Whisper transcription (small.en, medium, large models), audio format conversion via ffmpeg, CDR parsing for Twilio / Bandwidth / Cisco / generic RFC-3261, E.164 normalization, DNC scrubbing, silence-trim, loudness-normalize (EBU R128), voicemail-split (silence-based), IVR-prompt-render (SSML → audio via
edge-tts), phone-number-country-lookup, area-code demographics, hold-music-prep (loop-friendly wave), waveform-thumbnail, per-call log rollup, sudo-SMS queue (write-to-file, operator delivers). - 400 voicemails transcribed in ~12 minutes on a 40-core workstation with an A10M GPU (about 4 hours serial).
- Zero cloud calls. Whisper runs locally (whisper.cpp binary or Python's whisper package if you install it separately). Your audio never leaves your machine.
Every claim is testable on your own machine using the sample plans in the download.
Under the hood — the honest technical picture
The recipes wrap standard voice-adjacent tools: whisper / whisper.cpp, ffmpeg, and pure-Python parsers for CDR and phone-number formats. E.164 normalization is a hand-written parser (~80 lines) that avoids depending on phonenumbers. Sudo-SMS is a filesystem queue — recipe writes messages to a queue folder; a separate operator process (or a human) actually sends them.
- Whisper runs locally. Audio never leaves your machine. Whisper.cpp binary works fine; the Python whisper package works if you install it (it's the one exception to the “standard library only” rule — and only if you use the Whisper recipes).
- CDR parsers are format-specific. Twilio / Bandwidth / Cisco / generic RFC-3261 each have a small parser function. Add your own by copying an existing one.
- E.164 parsing is deterministic and readable. ~80 lines, handles US / Canada / international / extension.
- DNC scrubbing works from a local copy of the registry. You download the registry, the recipe scrubs against it locally — your call list never has to go online.
- Sudo-SMS is a queue, not a sender. Recipe writes intended messages to a folder; you (or an operator process on your carrier) actually deliver them. Cleaner separation and no Twilio dependency.
How it compares to the manual way
| Aspect | The typical alternative | This product |
|---|---|---|
| Transcribe 400 voicemails | Type them out by hand, or pay a cloud transcription service per minute | Run one recipe overnight, all transcribed locally |
| Scrub a call list against DNC | Web form, 10 numbers at a time | Local scrub against downloaded registry, 20,000 numbers in seconds |
| Parse today's carrier CDR | Excel + manual grouping | One recipe, per-caller summary as CSV |
| Normalize phone numbers | Excel formula gymnastics | One recipe, E.164 output |
What you need
- Windows / macOS / Linux workstation.
- Python 3.8+ for the recipe file. Standard library only — except the Whisper recipes need either the
whisper.cppbinary or thewhisperPython package (both external). - ffmpeg for audio-format recipes. Standard install.
- WonderPhi Compute to run the generated plan in parallel.
What's in the download
wpp-recipe.py— the recipe generator (Python 3.8+, standard library only)README.md— per-recipe reference for humansINDEX.md— alphabetical list of every recipe with a one-line descriptionsamples/— ready-to-run demo input files for every recipeLICENSE.txt— perpetual single-user license, personalized with your name and order number at delivery
Common questions
Q. Does it upload my audio anywhere?
No. Whisper transcription runs locally. Your voicemails never leave your machine.
Q. Which Whisper model should I use?
For English-only voicemail, small.en is fast and good enough. For accented speech or mixed language, medium or large gives noticeably better accuracy at a real time-cost. The recipe accepts --model to pick.
Q. How is 'sudo-SMS' different from Twilio?
Sudo-SMS is not an SMS gateway. It's a queue — the recipe writes intended messages to a folder as text files. A separate operator (usually you) picks them up and sends them via whatever mechanism you already have (your carrier, your phone, a Twilio account you own). It's the honest / no-vendor-lock-in way to plan a batch of outbound messages.
Q. Does it work with any carrier's CDR export?
Recipes ship with parsers for Twilio, Bandwidth, Cisco, and generic RFC-3261 CDR formats. If your carrier's format isn't covered, the parser functions are ~40 lines each — copy one and adapt.
Glossary — every term used above, in plain English
- Recipe
- One of the ready-made plan-file templates in this library. Each recipe knows how to do one job across many inputs.
- Plan file (
*.wpc.txt) - A plain text file with one shell command per line that WonderPhi Compute reads and executes across every CPU core in parallel.
- WonderPhi Compute
- The parallel runtime this recipe library is designed for. Sold separately, or bundled in the Complete Bundle.
- Shell command
- Any line you'd normally type into a terminal / Command Prompt / PowerShell prompt.
- Standard library only
- The recipe uses only what ships with Python 3.8+ — no
pip install, no external packages. - Perpetual license
- Buy once, use forever. No renewals, no expiration, no update check.
- CDR (Call Detail Record)
- The per-call log your phone carrier keeps: who called whom, when, for how long, and at what cost.
- Whisper
- OpenAI's open-source speech-to-text model. Runs locally on CPU or GPU. Good enough for voicemail-quality transcription.
- E.164
- The international standard phone-number format:
+country-code, then the number, no punctuation.+17707767023is E.164 for a US number. - DNC (Do Not Call registry)
- The US registry of phone numbers that have opted out of telemarketing calls. Businesses must scrub call lists against it or face FCC fines.
- IVR / SSML
- Interactive Voice Response menus (the “press 1 for sales” systems). SSML is Speech Synthesis Markup Language — the standard for specifying pronunciation, pauses, and prosody in TTS.
- EBU R128 / LUFS
- European Broadcasting Union recommendation for consistent audio loudness. LUFS is the unit.
License & support
Retail single-user license, perpetual, non-transferable. Every download's LICENSE.txt is personalized at checkout with your name, email, and order number. Full terms live inside your download — short version: install on any machine you personally own and operate, don't redistribute, don't run it as a paid service for third parties (that needs the Industrial variant), and everything else is fine.
Support: email chris@cri-one.com. Because WonderPhi Phone — 28 batch recipes for telephony & voice operations never changes, there's nothing to "support" in the update sense — but if something isn't behaving the way you expect, write and I'll help.
USA-only sales for retail.