Composable CLI for the Conversion Tools API.
Single-binary client. Pipe-friendly. Multi-region. MIT licensed.
# Convert a file
ctio convert -t json_to_excel data.json out.xlsx
# Pipe through stdin/stdout
cat data.json | ctio convert -t json_to_excel - - > out.xlsx
# Converters that take more than one file get the rest with --file
ctio convert -t validate_xml_xsd data.xml report.txt --file schema.xsd
# Compose
ctio parse invoice.pdf --schema fields.json \
| jq '.line_items' \
| ctio convert -t json_to_excel - out.xlsxWindows (scoop):
scoop bucket add conversiontools https://github.com/conversiontools/scoop-bucket
scoop install ctiomacOS and Linux (Homebrew):
brew install conversiontools/tap/ctioAny platform, direct download: grab the binary for your platform from
GitHub Releases
(ctio-darwin-arm64, ctio-darwin-x64, ctio-linux-x64, ctio-windows-x64.exe),
rename it to ctio (ctio.exe on Windows) and put it on your PATH:
chmod +x ctio
# macOS only: the binaries are not notarized, so clear the download quarantine once
xattr -d com.apple.quarantine ctioFrom source (needs Bun):
git clone https://github.com/conversiontools/ctio.git
cd ctio
bun install
bun run dev -- convert -t json_to_excel data.json out.xlsxGet your API token at https://conversiontools.io/profile → "API tokens".
ctio auth login # paste token, validates against the API
ctio auth status # shows active profile + masked tokenToken resolution order: --token flag > CT_API_TOKEN env > --profile <name> > active profile.
ctio list --from pdf # what converts from PDF
ctio list --to wav --format pretty # what produces WAV
ctio describe xml_to_csv --format pretty # options, allowed values, defaultsPull structured data out of a PDF, scan or photo with Parse.
Parse uses its own API key: create one at parse.conversiontools.io and pass it as
CT_PARSE_TOKEN (or --parse-token).
ctio parse invoice.pdf # schema-less: Parse picks the fields
ctio parse contract.pdf --schema fields.json -o out.json # just the fields you list
ctio parse --id <extraction_id> # fetch an earlier extraction- Fields.
fields.jsonis a JSON array of fields, or an object with afieldsarray (the shape Parse exports), e.g.[{"name": "total", "type": "number"}, {"name": "due_date", "type": "date"}].--schema-iduses a schema saved in Parse instead. - Waiting. ctio waits until the extraction finishes and prints the data. A long document
can take a few minutes;
--format prettyshows the progress.--timeout <sec>caps the wait (default 1800,0= no cap). If it runs out, ctio prints the extraction id, andctio parse --id <id>picks it up later. - Failures. A failed extraction exits non-zero with the reason. ctio never prints
nullfor a result it did not get. - Repeats. The same document with the same fields returns the stored result, and no pages are charged.
Narrate a text, Markdown, Word or PDF file:
ctio convert -t ai_text_to_mp3 script.txt narration.mp3 --option voice_tts=en_paul_neutral
ctio convert -t ai_pdf_to_mp3 handbook.pdf handbook.wav --option voice_tts=gb_jane_neutral-
Voices.
ctio voice --format pretty(orctio voices) lists the 66 built-in voices in 21 languages (voice_478Sheila,en_paul_neutral,gb_jane_neutral,fr_marie_neutral,voice_3133Haruto, ...) and the voices you cloned (see Clone your voice). A cloned voice works from the account that cloned it. An unknown voice is an error, never a silent fallback. -
Language. Each built-in voice lists the
language_ttscode it speaks (jafor Haruto). Set--option language_tts=<code>to the language your text is written in, and pick a voice that speaks it.ctio describe ai_text_to_mp3lists all 38 codes. -
Format. MP3 or WAV, taken from the output file name (
narration.wav), or set with--option output_tts=wav. Writing to stdout (-) gives MP3 unlessoutput_ttssays otherwise. -
Names.
ai_text_to_speech,ai_docx_to_speech,ai_pdf_to_speechandai_md_to_speechwork as converter names too. -
Sound. The audio comes out at spoken-word loudness (about -16 LUFS), 24 kHz mono in both formats, so it can go straight into a video. MP3s report their true duration.
Text to speech uses AI credits (10 per 1,000 characters).
Record about 10 seconds of clear speech (one speaker, a quiet room), then clone it and narrate in it:
ctio voice clone my-voice.wav --name "My narrator" --tone calm \
--consent "I confirm this is my own voice, or I have the speaker's explicit permission to clone it."
ctio convert -t ai_text_to_mp3 script.txt narration.mp3 --option voice_tts=<voice_id>
ctio voice delete <voice_id> --yes- Consent.
--consenttakes the consent statement word for word (ctio voiceprints it asconsent_statement). Without it ctio stops before the sample leaves your machine. Clone only your own voice, or a voice whose speaker gave you explicit permission. - Sample. MP3, WAV, M4A, FLAC, OGG or WebM. 3 seconds minimum, about 10 seconds works best. Up to 512 KB on the free Voice plan, 2 MB on paid plans.
- Tone. A label for the sample's mood:
normal,excited,cheerful,confident,curious,calm,sad,angryorfrustrated. The clone copies the mood and pace of the recording itself, so record in the mood you want to hear. - Slots. 1 cloned voice on the free Voice plan, 5 on Starter, 20 on Pro. Deleting a voice frees its slot.
- Voices cloned with ctio show up on the voice.conversiontools.io
dashboard, and voices cloned there show up in
ctio voice. You can also record a sample right in the dashboard.
MIT