As checked 2026-10-25: Deepgram’s official pricing page lists Aura-2 text-to-speech under a per-1,000-character billing unit, distinct from its per-minute voice-agent offerings. This is a dated economics guide, not a permanent price guarantee or a personal hands-on benchmark.
Scope: global English reference; USD list prices where shown; local eligibility, currency, taxes, negotiated terms and model release dates require separate confirmation. Dynamic source-page prices have a check date, not a verified initial effective date.
What is actually metered
Deepgram’s official pricing page lists Aura-2 text-to-speech under a per-1,000-character billing unit, distinct from its per-minute voice-agent offerings. In the October 25, 2026 observation, Aura-2 appears at $0.030 per 1,000 characters on pay-as-you-go and $0.027 per 1,000 on the Growth schedule. These are list-rate observations for a specified product; they do not establish what a negotiated enterprise or cloud-reseller arrangement charges.
Character count is not audio length
A multilingual narration script might contain the same number of characters in two languages while resulting in very different speech duration. To forecast a character-priced TTS workflow, count billable input characters according to the provider’s documented rules rather than estimating from finished minutes. For a hypothetical 500,000 billable characters, the quoted pay-as-you-go Aura-2 arithmetic is 500 × $0.030 = $15. That excludes other services and assumes the entire request matches the quoted SKU and tier.
Beware blended voice-agent rates
A separate Voice Agent product can combine speech recognition, language processing and voice generation under its own session-minute meter. A buyer cannot add a TTS per-character charge to a bundled agent rate unless both services are explicitly billed. Likewise, an external telephony network or a bring-your-own-model configuration can produce additional costs or remove included components. Model the exact architecture and contract.
Operational variables
Record voice/model ID, text normalization, language, connection retries, truncation, audio quality and whether a request succeeded. Character rounding or minimum billing units should come from the live developer documentation or an invoice if not disclosed on the marketing price page. The public page’s promotional credits or limited-time offers should never replace the recurring base rate in a forecast.
Best use of this price reference
This unit is useful when comparing the cost of a stable narration volume from the same input texts. It does not establish speech quality, pronunciation or licensing rights. For real-time agent work, compare end-to-end turn latency and recognition, generation and transport costs under the appropriate separate offer.
What belongs in the character meter
Take the text after normalization, template substitutions and any provider-specific markup handling, then count according to the API’s published billed-character definition. A rough count from the original manuscript may miss dynamically inserted names, captions or translated passages. Track the character total returned by the service where available, the exact voice version and whether the call generated a usable result. If billing treatment for punctuation, markup or failed calls is undocumented, leave the rule unresolved and validate it with an account invoice or support reply.
Narration production planning
An editor producing a multilingual series should forecast each language as its own character workload. Speech duration, naturalness and revision frequency should be evaluated separately. Suppose 300,000 billable characters are processed at the published pay-as-you-go Aura-2 rate: 300 × $0.030 yields $9 in that meter, but additional voices, live-agent components or editing service fees remain outside the subtotal. Different vendors may charge by generated minutes instead, so normalizing by accepted finished audio is necessary for an operational comparison.
Decision checklist
- Identify the specific plan or API endpoint, model version, region and currency before pricing a job.
- Measure input, output, caching, generated media and failed-job units separately.
- Check the official pricing page again before procurement, especially if the SKU has no effective-date notice.
- Keep source evidence and a timestamped price ID; do not treat unknowns as zero.
Pricing evidence boundary
This cost discussion concerns Deepgram. Reproduce a quote only with its exact model or plan identifier, chargeable unit, billing region, processing mode and timestamp; a changing vendor rate card is not a permanent price guarantee. If any one of these dimensions is missing, keep the associated rate and worked scenario under editorial review instead of filling the gap with an estimate.
Sources & Further Reading
- Deepgram Aura-2 per-character TTS list tariffs and other minutes-based channels — official first-party source, accessed 2026-10-25; original effective date not specified in source.