TTSMaker

The free web converter lists a weekly allowance of 20,000 characters, with some voices treated separately as unlimited. This is a character budget, not a promise of a fixed number of minutes. Two scripts of identical length can have different playback durations after voice, speed and pause choices change.

The pricing comparison lists 500 characters per Free conversion, while the live converter can display a different maximum for the selected voice. Keep those values separate from the weekly allowance. Paste a short passage first, read the remaining-character indicator and stop if the requested conversion exceeds the visible limit; a large weekly balance cannot override a smaller request limit.

TTSMaker permits lawful commercial use of synthesized speech and says source attribution is optional. The practical deliverable is a downloaded audio file, rather than an audition that can only be played inside the service. Save the final script with the generation date and the terms link so the recipient knows which version was produced.

Use text and other material you have permission to process. A speech-generation licence does not supply permission for a borrowed script, music track or a third party’s identity. Before adding background music, check that material separately and keep an unaccompanied voice file available for editing; do not turn the provider’s commercial-use statement into a blanket clearance of the entire production.

Editorial assessment

Begin with Free for a short, individually downloaded voiceover that fits the chosen voice. Lite addresses longer web conversions but remains outside the documented API route. A developer who needs unattended requests should evaluate the eligible PRO or Studio subscription instead of assuming that a paid web plan always includes programming access.

The API returns a temporary audio URL rather than an audio stream, and the documentation gives that URL a 24-hour lifetime. Store the downloaded result in your own authorized destination before passing it to downstream work. The API charges voices in regular character mode, so an unlimited voice badge in the website must not be used to predict unlimited API usage.

Choose the language and voice before adjusting speed, pitch, volume and pauses. Record the visible settings with each saved script. For a multi-part narration, split at complete sentences and keep a short repeated calibration sentence outside the final deliverable; listening to it across files helps reveal an accidental voice or speed change.

The user guide describes the text-to-MP3 workflow, but it is older than the current controls. Use the options actually available in the account rather than expecting every screenshot to match. When a name sounds wrong, revise one spelling or punctuation choice at a time, retain the previous audio, and compare the entire sentence for a change in meaning.

The Free interface describes a 30-minute history window, while the privacy page refers to text deletion after 60 minutes based on settings and service terms describe storage up to 24 hours. These statements cover different surfaces and are not a precise retention guarantee for every request. Download promptly and avoid confidential material until the applicable handling is clear.

Keep a local folder with the approved script, voice identifier, chosen settings, final audio and correction notes. A temporary history link is not the project archive. Reopen the actual downloaded MP3 in another player and inspect the start and end; a successful generation message alone cannot confirm that the file your recipient receives is intact.

1. Keep the sentence unchanged for the first conversion and choose MP3 if offered.
2. Download immediately, then open the saved file in a separate player.
3. Listen for the negation, collection time, Q7, B12 and the name Mara.
4. If one item is unclear, change only that pronunciation cue and generate one correction.
5. Record both character deductions and keep the two files for comparison.

Accept this brief only if the saved audio can be reopened and preserves all five meaning-sensitive items. Fail on truncation, a lost negation, an unavailable download or an unwanted paid step. Leave actual observations blank until the exercise is run; the quota calculation above is planning arithmetic, not an observed production yield.

Compare the access conditions in ElevenLabs and Murf, then use the free-output selection guide to distinguish an audition from a deliverable file.