Speech-to-Text & TTS Cost Calculator - Compare API Pricing

Speech-to-text cost

Simple and Compare modes, for a speech-to-text model

Cost

the cost of one run

U

the billed usage, in minutes: the audio length, taken up to a whole number of billing increments

R

the price per minute of audio

Rates are in US dollars per minute of audio or per 1,000 characters, as each provider’s pricing page states them, or an estimate marked “est.” where the provider bills by token. This panel carries no prices: the model details under the model picker show the rate in use and the date of the last check.

Source: Standard formula; no single publisher. Rates are each provider’s own, such as AWS’s [1] and Google Cloud’s first-tier rate [3].

Text-to-speech cost

Simple and Compare modes, for a text-to-speech model

Cost

the cost of one run

U

the billed usage, in characters, taken up to a whole number of billing increments

R

the price per 1,000 characters; the Neural rate when Voice Quality is Neural

Google Text-to-Speech and AWS Polly list both a standard and a Neural voice; Voice Quality picks which rate applies.

Source: Standard formula; no single publisher.

Options: realtime, speaker labels and profanity filter

Simple and Compare modes, for a speech-to-text model with an option that adds a charge

S

the surcharges of the ticked options, per run

Each option is priced the way the model’s provider prices it: the streaming rate less the file rate for each billed minute, an add-on per hour of audio, or nothing where the option is included. Where no price is on record, S is the percentage you enter of the base cost; several percentages add rather than compound. They start at 25 for realtime, 12 for speaker labels and 8 for the profanity filter. A model that does not offer a ticked option is ranked last in Compare. Speaker labels and the profanity filter apply to speech-to-text only; on a text-to-speech model, Realtime is always your own percentage.

Source: Each provider’s own option pricing, such as AWS’s streaming rate [1] and AssemblyAI’s speaker-label add-on per hour [2]. The starting percentages are CalcTrail’s own placeholders; no publisher.

Billed usage

Simple and Compare modes, with a billing increment set

U

the billed usage

Increment

the block that usage is billed in, in seconds for speech-to-text and characters for text-to-speech

ceil rounds up to a whole number. With no increment typed, usage is rounded up to the model’s own block: one second for speech-to-text and one character for text-to-speech. A per-file minimum, such as Rev AI’s 15 seconds, is a different rule and is not applied.

Source: Each provider’s billing unit, such as AWS’s one-second increments [1] and Rev AI’s per-file minimum [4]; an increment you set is your own assumption.

Monthly cost

Simple and Compare modes

Monthly Cost

the cost of a month of runs

Cost

the cost of one run

Runs per Month

a whole number of runs; a blank entry counts as one

Source: Standard formula; no single publisher.

Ranking in Compare

Compare mode

Every ticked model of the same type is priced on the same usage, runs and options, then ranked by cost per run. Each uses its own rate and billing increment unless you set one. A custom rate applies to the selected model only.

Source: CalcTrail’s own ranking rule; no publisher.