Hub · More Tools
Tonwerk
90 published endpoint URLs. Methods, prices, inputs and original documentation on every detail page.
ton.halowerk.com · Catalog updated 2026-11-17 UTC
Endpoints
61–90 of 90- Tonwerk0.02 USDC
script adherence
POSThttps://ton.halowerk.com/v1/script-adherenceEach script step (with optional variants) is searched in the agent's turns with a fuzzy phrase matcher: content words (Snowball stop words removed) are matched exactly, by light German/English suffix stemming or by Levenshtein similarity ≥
- Tonwerk0.01 USDC
scrub sprites
POSThttps://ton.halowerk.com/v1/scrub-spritesTakes one thumbnail every interval_s (default: 1, 2, 5, 10, 15, 30 or 60 s so that at most 120 thumbnails result), scales it to the requested width with the display aspect ratio, tiles the thumbnails into sheets of columns x rows with ffmpe
- Tonwerk0.02 USDC
sign overlay space
POSThttps://ton.halowerk.com/v1/sign-overlay-spaceSamples the video at sample_fps on a 320 px wide grey copy. For each candidate corner a box of area_share of the picture (default 1/6, the minimum signer size Ofcom concluded in its review of signing on television) with aspect box_aspect an
- Tonwerk0.02 USDC
slides from video
POSThttps://ton.halowerk.com/v1/slides-from-videoSamples the video at sample_fps on a 160 px wide grey image, splits it into runs without change (share of pixels changing by more than 24 grey levels at most change_threshold, against the previous sample and the run start) and keeps runs la
- Tonwerk0.005 USDC
speech music segment
POSThttps://ton.halowerk.com/v1/speech-music-segmentDecodes up to 10 minutes (mono, 16 kHz). Frame features (64 ms, 20 ms hop): energy, spectral flatness 100–6000 Hz, periodicity as Boersma (1993) window-corrected autocorrelation (70–500 Hz) and zero-crossing rate. Per 1 s window (0.5 s hop)
- Tonwerk0.02 USDC
spoken language id
POSThttps://ton.halowerk.com/v1/spoken-language-idSilero VAD v6.2 finds the speech; sections separated by at least 0.6 s of silence are scored separately (closest sections are merged until at most max_sections remain, sections shorter than min_section_s are listed but not scored). For the
- Tonwerk0.005 USDC
ssml validate
POSThttps://ton.halowerk.com/v1/ssml-validateParses the SSML (well-formedness, no DTD) and checks it against W3C SSML 1.1: <speak> with version="1.1", xml:lang and namespace; the content model of every element (e.g. no <p> in <s>, <desc> only in <audio>, lexicon/meta first); required
- Tonwerk0.02 USDC
stem split
POSThttps://ton.halowerk.com/v1/stem-splitDecodes up to 45 s to 44.1 kHz stereo (mono is duplicated) and runs Deezer Spleeter 2stems (U-Net, MIT licence) as ONNX: the export made with the sherpa-onnx conversion scripts (Apache-2.0), int8, repository csukuangfj/sherpa-onnx-spleeter-
- Tonwerk0.02 USDC
stream clip secure
POSThttps://ton.halowerk.com/v1/stream-clip-secureReads an HLS master or media playlist (RFC 8216) from a public URL, picks the lowest or highest bandwidth variant, selects the media segments covering `seconds` at the live edge (newest published segments) or from start_offset_s, and fetche
- Tonwerk0.01 USDC
stream loudness log
POSThttps://ton.halowerk.com/v1/stream-loudness-logMeasures a recording of up to 600 s (for example a stream dump) with the ffmpeg ebur128 filter (ITU-R BS.1770 K-weighting, 400 ms momentary and 3 s short-term loudness every 100 ms, true peak by oversampling). For each interval of interval_
- Tonwerk0.01 USDC
stt quality forecast
POSThttps://ton.halowerk.com/v1/stt-quality-forecastSilero VAD v6.2 separates speech from pauses. SNR: active speech power (20-ms frames detected as speech and within 30 dB of the loudest) minus noise power, against the noise power of clear pauses. Reverberation: free decays after level peak
- Tonwerk0.02 USDC
stt vocab boost
POSThttps://ton.halowerk.com/v1/stt-vocab-boostRuns OpenAI Whisper base (MIT; ONNX export Xenova/whisper-base, pinned revision, CPU) twice on the same encoded audio: once plain, once with the vocabulary as a decoder prompt ("Glossary: term, term.", the initial_prompt mechanism of openai
- Tonwerk0.01 USDC
transcript compare
POSThttps://ton.halowerk.com/v1/transcript-compareBoth transcripts (plain text or timed segments) are split into words, normalised per options (NFKC, case folding, punctuation removed) and aligned with a word-level Levenshtein alignment. Consecutive non-matching words form one difference r
- Tonwerk0.003 USDC
transcript pages
POSThttps://ton.halowerk.com/v1/transcript-pagesTakes episodes inline or from an RSS feed whose items link transcripts via podcast:transcript (up to four transcripts are fetched, preferring VTT, JSON, SRT, HTML). Parses WebVTT voice spans, SRT cards with "Name:" speaker prefixes, word-le
- Tonwerk0.005 USDC
true bandwidth
POSThttps://ton.halowerk.com/v1/true-bandwidthReads codec, sample rate and bit depth with ffprobe, decodes up to 120 s at the native rate (mono mix) and computes a Welch long-term spectrum (Hann, 50 % overlap, about 6 Hz resolution) smoothed over 200 Hz. A cliff candidate is the highes
- Tonwerk0.01 USDC
tts ab compare
POSThttps://ton.halowerk.com/v1/tts-ab-compareDecodes both clips (base64 or URL, at most 10 MiB and 60 s each) with ffmpeg 6.1 and measures each on 10 ms frames: speech segments above the silence threshold (blips under 30 ms dropped, gaps shorter than min_pause_ms bridged), leading/tra
- Tonwerk0.005 USDC
tts text normalize de
POSThttps://ton.halowerk.com/v1/tts-text-normalize-deRule-based German text normalisation for TTS front ends. Detects ISO and German numeric dates (12.11.2026, 3. Oktober), times (14:30 Uhr), amounts in EUR/USD/CHF/GBP (12,50 €), numbers with units (km/h, kWh, °C, %, Mio.), paragraphs (§ 5a),
- Tonwerk0.005 USDC
tuning reference
POSThttps://ton.halowerk.com/v1/tuning-referenceDecodes up to 10 minutes (mono, 22.05 kHz). Spectral peaks between 80 and 2000 Hz (about 2.7 Hz bins, 75 % overlap) are located with parabolic interpolation on the log power; each peak gives its deviation from the equal-tempered grid around
- Tonwerk0.01 USDC
turn latency profile
POSThttps://ton.halowerk.com/v1/turn-latency-profileTakes a stereo call recording (up to 300 s, user and agent on separate channels), finds speech per channel with Silero VAD v6.2 (MIT, pinned) and joins user speech separated by less than merge_user_pause_ms into turns. For each user turn th
- Tonwerk0.01 USDC
vad events
POSThttps://ton.halowerk.com/v1/vad-eventsRuns Silero VAD v6.2 (MIT, pinned ONNX model, 32 ms frames at 16 kHz) over a recording of up to 300 s and replays it the way the VADIterator of snakers4/silero-vad emits events in a live stream: a start event when the speech probability cro
- Tonwerk0.02 USDC
verbatim protocol
POSThttps://ton.halowerk.com/v1/verbatim-protocolSpeakers come from audio channels: with a two-channel recording (e.g. a call or an interview with one microphone per person) each channel is one speaker; identical channels or a mono file give one speaker (no voice-based diarisation). Per c
- Tonwerk0.015 USDC
video denoise
POSThttps://ton.halowerk.com/v1/video-denoiseFilters with ffmpeg 6.1: hqdn3d (spatio-temporal, default), atadenoise (adaptive temporal averaging, keeps edges, needs a mostly static camera) or nlmeans (non-local means, best detail but slow, therefore limited to short low-resolution cli
- Tonwerk0.02 USDC
video dup detect
POSThttps://ton.halowerk.com/v1/video-dup-detectBoth videos (up to 300 s each) are sampled at sample_fps from their start, each sample is reduced to 64×64 grey and hashed with a 64-bit DCT perceptual hash (pHash method; aspect ratio ignored, so rescaling does not matter). Flat samples (s
- Tonwerk0.02 USDC
video quality compare
POSThttps://ton.halowerk.com/v1/video-quality-compareBoth videos are decoded from start_s for duration_s (at most 60 s), converted to 8-bit YUV 4:2:0 and aligned frame by frame at the frame rate of the reference; the distorted video is scaled to the reference size with a bicubic filter when t
- Tonwerk0.02 USDC
video stabilize
POSThttps://ton.halowerk.com/v1/video-stabilizePass 1 runs ffmpeg 6.1 vidstabdetect (vid.stab 1.1, local motion fields), pass 2 vidstabtransform with the chosen smoothing window (frames), border handling (static optimal zoom, adaptive zoom, black or kept borders) or tripod mode, followe
- Tonwerk0.03 USDC
video text ocr
POSThttps://ton.halowerk.com/v1/video-text-ocrSamples the video at sample_fps; only samples whose 160 px grey image differs from the last OCR frame (more than 0.3 % of pixels changing by more than 24 levels) are read again, at most max_ocr_frames per call (the largest changes win). The
- Tonwerk0.005 USDC
vocal range
POSThttps://ton.halowerk.com/v1/vocal-rangeDecodes up to 5 minutes (mono, 16 kHz) and tracks the fundamental frequency with YIN (de Cheveigné & Kawahara 2002: cumulative mean normalised difference, absolute threshold 0.15, parabolic refinement), 60–1500 Hz, 10 ms hop. Frames count a
- Tonwerk0.005 USDC
voice agent eval
POSThttps://ton.halowerk.com/v1/voice-agent-evalTakes the turns of a voice-agent conversation (speaker, text, optional start/end times) and your goal criteria (required phrases per speaker, success and failure phrases). Without any language model it detects, with German and English phras
- Tonwerk0.02 USDC
voicemail triage
POSThttps://ton.halowerk.com/v1/voicemail-triageRule-based triage with disclosed German/English cue patterns: seven concern categories (billing, appointment, order/delivery, technical issue, complaint, contract, information request; most hits wins, ties by first mention), urgency (explic
- Tonwerk0.01 USDC
wer benchmark
POSThttps://ton.halowerk.com/v1/wer-benchmarkTexts are normalised (NFKC, optional lower-casing and punctuation removal, whitespace collapsed) and aligned with a word-level Levenshtein alignment (unit costs). WER = (S + D + I) / N with N reference words; MER = (S + D + I) / (H + S + D