Best ElevenLabs Alternatives in 2026

If you are shopping for ElevenLabs alternatives in 2026, start with Cartesia Sonic Pro at $5/month or Hume Octave Creator at $14/month. Cartesia Sonic is the closest drop-in replacement for the ElevenLabs Studio workflow at a fraction of the cost. Hume Octave is the better pick when empathy and voice design matter as much as raw voice quality. This guide compares nine verified alternatives, with current pricing pulled from each vendor’s official pricing page in July 2026.

ElevenLabs alternatives at a glance

Alternative	Best for	Free plan	Starting price	Main advantage	Main limitation
Cartesia Sonic	Real-time voice and voice agents	Yes	$5/month (Pro)	Sub-100 ms latency, instant voice cloning	Smaller voice library than ElevenLabs
Hume Octave	Empathic voice and speech-to-speech	Yes	$3/month (Starter)	LLM-based voice design, EVI 3 bundled	Per-character overage on lower tiers
OpenAI gpt-realtime	Voice agents inside ChatGPT	Limited	$20/month (Plus)	Native speech-to-speech on Realtime API	Per-token pricing harder to budget
Google Cloud TTS	Enterprise scale and Vertex AI	Yes	$30 per 1M chars (Chirp 3: HD)	Gemini-TTS family, multi-region	Emotional range trails ElevenLabs v3
Azure Speech	Microsoft 365 and Azure shops	Yes	Pay-as-you-go per char	Voice Live API, 600+ voices	Best features behind commitment tiers
Speechify	Individual readers and students	Yes	$29/month (Premium)	1,000+ voices, voice typing, AI podcasts	Curated library, not API-first
Amazon Polly	AWS-native apps and IVR	Yes	$4 per 1M chars (Standard)	Predictable per-character billing, AWS IAM	Generative voices less expressive
Resemble AI	Voice cloning and AI security	Yes	$0.0005/sec TTS	Rapid and Pro cloning, watermarking	Pay-as-you-go can add up
Murf AI	Enterprise content production	Yes	$0.01/min (Falcon API)	200+ voices, sub-55 ms Falcon API	Subscription pricing is sales-led
Kokoro TTS	Self-hosting and zero per-char fees	Yes	$0 self-hosted	Apache-2.0 weights, run anywhere	You operate the hosting and safety stack

Why look for an ElevenLabs alternative?

ElevenLabs still owns the most expressive consumer TTS, the deepest voice-cloning story, and the broadest community voice library on the market. Eleven v3 went generally available on February 2, 2026 with 70+ languages and a 3,000 character limit per request, Eleven Music v2 launched May 26, 2026 with section-by-section composition and inpainting, and Dubbing v2 released May 28, 2026 with emotion preservation across languages.

Three practical reasons push buyers off ElevenLabs anyway. First, the credit pool is shared across TTS, music, sound effects, voice changer, and dubbing, so heavy music or dubbing use drains the budget you set aside for narration. Second, the free tier does not allow commercial use, the first commercial tier is Starter at $6/month, and free credits do not roll over while paid credits only roll over for two months. Third, per-character cost on Eleven v3 and Multilingual v2 is higher than cloud-provider TTS at scale, especially when you generate millions of characters of audiobook narration or IVR prompts.

If your core workflow is Eleven v3 emotional range, Professional Voice Cloning, Eleven Music v2, or Dubbing v2, ElevenLabs is still the right default. The moment your priority is real-time voice agents, predictable per-character billing, or self-hosted inference, at least one alternative will give you more for less.

How I chose these alternatives

I picked nine alternatives that overlap with ElevenLabs’ core jobs: expressive text-to-speech, voice cloning, conversational agents, dubbing, and audio APIs. Each entry below is verified against an official pricing page or documentation page captured on 2026-07-11, cross-checked against ElevenLabs’ own pricing page, model documentation, and blog announcements. I excluded niche consumer read-aloud apps and pure music generators because they do not match the voice-platform workflow that ElevenLabs owns.

1. Cartesia Sonic — Best overall

Cartesia Sonic is the strongest ElevenLabs alternative for developers who care about real-time latency, voice cloning, and voice agents. The Pro plan at $5/month includes a commercial license, instant voice cloning, and 100,000 credits per month, which is roughly 133 TTS minutes on Sonic-3.5. The Sonic-3.5 model ships sub-100 ms latency that fits real-time voice agents, and the Cartesia Line voice-agent product shares the same credit pool.

• What it does. Generates speech with Sonic-3.5, transcribes audio with Ink-2, runs voice agents with Line, and supports instant and Pro voice cloning.
• Best for. Developers building real-time voice agents or chat agents who want voice cloning and TTS in one stack.
• Why valid. The Cartesia pricing page documents Free 20K credits, Pro 100K credits at $5/month, Startup 1.25M credits at $49/month, Scale 8M credits at $299/month, and Enterprise custom with DPAs, BAAs, and SSO. Pro voice cloning unlocks on Startup tier and above.
• Strengths. Sub-100 ms Sonic-3.5 latency, commercial license from $5/month, voice-agent slots included on every paid tier, real-time WebSocket API.
• Limitations. Smaller community voice library than ElevenLabs’ 10,000+ marketplace, and Pro voice cloning is gated above Pro.
• Pricing. Free; Pro $5/month; Startup $49/month; Scale $299/month; Enterprise custom.
• Free/trial. Yes, 20,000 credits per month with TTS and STT.
• Migration. Re-record voice samples for instant cloning on Pro, then port any workflow built around ElevenLabs Studio into Cartesia Studio.

Verdict

Pick Cartesia Sonic Pro if you want ElevenLabs’ Studio workflow at $5/month instead of $22/month. Pair the choice with the broader ElevenLabs comparison in 9 AI voice tools that created professional audio content.

2. Hume Octave — Best for empathic voice agents

Hume Octave is the best ElevenLabs alternative when empathy and voice design matter as much as raw voice quality. Octave is an LLM-based TTS that reasons about emotion and pacing, and the same billing plan unlocks EVI 3 speech-to-speech agents for real-time voice conversations. Unlimited voice cloning is included on every paid tier, including Starter at $3/month.

• What it does. Generates speech with Octave, runs EVI 3 / EVI 4 mini speech-to-speech agents, supports unlimited voice cloning, and exposes the same API for both modalities.
• Best for. Voice agents, customer experience, and creators who want emotion-aware voice design.
• Why valid. The Hume AI pricing page lists Free with 10,000 TTS characters, Starter at $3/month with 30,000 characters and 40 EVI minutes, Creator at $7 first month then $14 with 140,000 characters and 200 EVI minutes, and Pro at $70/month with 1,000,000 characters and 1,200 EVI minutes.
• Strengths. LLM-based voice design, unlimited cloning on every paid tier, SOC 2 Type II, GDPR, and HIPAA on Business and above, voice conversion included on Pro and above.
• Limitations. Per-character overage on lower tiers ($0.05–$0.15 per 1,000 characters), and voice library is smaller than ElevenLabs’ community marketplace.
• Pricing. Free; Starter $3/mo; Creator $7/mo first month then $14/mo; Pro $70/mo; Scale $200/mo; Business $500/mo; Enterprise custom.
• Free/trial. Yes, with 10,000 TTS characters and 5 EVI minutes.
• Migration. Re-export ElevenLabs voice samples, clone them on Hume, then port any conversational prompts into EVI 3.

Verdict

Pick Hume Octave Creator if empathy and reasoning about tone beat raw expressiveness for your use case. See best AI voice model: Microsoft MAI Voice 2 for context on the broader voice-model landscape.

3. OpenAI Text-to-Speech and Realtime — Best for ChatGPT stack

OpenAI’s Realtime API and ChatGPT Advanced Voice are the most natural ElevenLabs alternative when your voice workflow already runs on GPT-5 and you want speech-to-speech agents, TTS, and STT on the same platform. The gpt-realtime-2.1 model supports native audio input and output, and gpt-realtime-2.1-mini is the cost-sensitive variant for high-volume voice agents.

• What it does. Generates speech with TTS, transcribes with Whisper, runs native speech-to-speech agents with gpt-realtime, and exposes WebRTC, WebSocket, and SIP connection methods.
• Best for. Voice agents, ChatGPT users, and developers already wiring GPT-5 into their stack.
• Why valid. The OpenAI API pricing page lists gpt-realtime-2.1 audio at $32 per 1M input tokens and $64 per 1M output tokens, gpt-realtime-2.1-mini at $10/$20 per 1M tokens, gpt-realtime-translate at $0.034 per minute, and gpt-realtime-whisper at $0.017 per minute. ChatGPT Plus runs $20/month and Pro runs $200/month with 5× or 20× more usage.
• Strengths. Native speech-to-speech, mature agent SDK, single platform for TTS, STT, translation, and agent orchestration.
• Limitations. Per-token audio pricing is harder to budget than character-based plans, and there is no community voice library on the scale of ElevenLabs’ marketplace.
• Pricing. ChatGPT Free; Plus $20/month; Pro $200/month; gpt-realtime-2.1 audio $32/$64 per 1M tokens.
• Free/trial. Yes, ChatGPT free tier includes limited voice mode.
• Migration. Re-author ElevenLabs voice clone scripts as OpenAI voice prompts, then port any Conversational AI flow into the Agents SDK or Realtime API.

Verdict

Pick OpenAI Realtime if you want voice agents inside the broader ChatGPT and GPT-5 stack. The OpenAI developers blog covers voice-agent deployment patterns in detail.

4. Google Cloud Text-to-Speech — Best for enterprise scale

Google Cloud TTS is the strongest ElevenLabs alternative for enterprise scale, multi-region deployments, and Vertex AI integration. The Gemini-TTS family brings LLM-style prompt control to TTS, and Chirp 3: HD voices cover most general-purpose workloads at $30 per 1M characters with a 1M character free tier.

• What it does. Generates speech with the Gemini-TTS family, Chirp 3: HD voices, and legacy Standard, WaveNet, Neural2, Studio, and Polyglot voices, and exposes the same API across 220+ voices and 40+ languages.
• Best for. Enterprises running on Google Cloud, teams that need custom voices, and multilingual IVR.
• Why valid. The Google Cloud Text-to-Speech pricing page lists Gemini 2.5 Flash TTS at $0.50 input and $10 output per 1M tokens, Gemini 2.5 Pro TTS at $1/$20, Chirp 3: HD at $30 per 1M characters (1M free), Instant custom voice at $60 per 1M characters, and legacy tiers from $4 per 1M characters on WaveNet and Standard.
• Strengths. Gemini-TTS prompt control, 220+ voices across 40+ languages, free usage on legacy tiers up to 4M characters, deep Vertex AI integration.
• Limitations. Voice acting and emotional range trail ElevenLabs v3, and per-character billing gets expensive for casual hobbyists.
• Pricing. Gemini-TTS per 1M tokens; Chirp 3: HD $30 per 1M chars; legacy from $4 per 1M chars.
• Free/trial. Yes, up to 4M characters per month on WaveNet and Standard.
• Migration. Export ElevenLabs narration scripts as plain text, run them through the Gemini-TTS API with prompt-level emotion control, and replace voice clones with Chirp 3 Instant Custom Voice samples.

Verdict

Pick Google Cloud TTS if your stack already runs on Vertex AI or you need predictable multi-region deployment. See 9 AI voice tools that created professional audio content for additional voice-platform options.

5. Azure Speech in Foundry Tools — Best for Microsoft 365 shops

Azure Speech in Foundry Tools is the right ElevenLabs alternative when your team already lives inside Microsoft Azure or Microsoft 365 and you want TTS, STT, and voice-agent infrastructure in one stack. The Voice Live API unifies speech-to-text, LLM, and text-to-speech in one WebSocket, and the Azure Speech pricing page lists a Free tier with 0.5 million Neural characters per month.

• What it does. Synthesizes speech with Neural and Neural HD voices, supports Custom Neural Voice, Personal Voice, and avatar output, and exposes the Voice Live API for real-time voice agents.
• Best for. Microsoft 365 shops, Azure-native contact centers, and teams that need BAAs for HIPAA.
• Why valid. The Azure Speech in Foundry Tools pricing page lists a Free tier with 0.5 million Neural characters per month, pay-as-you-go per-character Neural and Neural HD rates, and the Voice Live API with Pro, Basic, and Lite tiers per 1M tokens. The Voice Live API documentation describes 600+ standard voices across 150+ locales and integration with GPT-Realtime, GPT-5, GPT-4.1, Phi, and other models.
• Strengths. Voice Live API unifies STT, LLM, and TTS in one WebSocket, 600+ standard voices, custom voice and avatar available, deep Azure compliance story.
• Limitations. Most compelling features sit behind Azure commitment tiers, and Personal Voice is a limited-access feature with eligibility requirements.
• Pricing. Free 0.5M Neural chars/month; paid per character; Voice Live Pro, Basic, and Lite per 1M tokens.
• Free/trial. Yes, the Free F0 tier includes 0.5 million Neural characters per month.
• Migration. Re-author ElevenLabs voice designs as Custom Neural Voice training jobs, then port any conversational agent flow into Voice Live API endpoints.

Verdict

Pick Azure Speech when your voice workflow needs to live inside Microsoft 365 or Azure. For context on MAI Voice, see Microsoft MAI Voice 2 review and Microsoft MAI Voice 2 pricing.

6. Speechify — Best for individual readers and students

Speechify is the best ElevenLabs alternative for personal reading, listening, and dictation rather than API work. The Premium plan at $29/month includes 1,000+ voices across 60+ languages, 5x listening speed, voice typing, AI podcasts, and a Voice AI Assistant.

• What it does. Reads articles, PDFs, docs, and books aloud, generates AI podcasts, and supports voice typing dictation inside the Speechify app.
• Best for. Students, professionals with reading-heavy workflows, and casual creators who want a polished reading app.
• Why valid. The Speechify pricing page lists a free tier with 10 robotic voices at 1.5x speed and Premium at $29/month with 1,000+ voices, 60+ languages, 5x listening speed, AI summaries, and integrated Voice Typing.
• Strengths. 1,000+ curated voices, voice typing and AI podcasts bundled, iOS, Android, Chrome, Edge, macOS, and Windows clients.
• Limitations. Curated voice library, not ElevenLabs’ 10,000+ community marketplace, and not as compelling for developers as API-first competitors.
• Pricing. Free; Premium $29/month; separate Speechify Studio and Speechify API pricing.
• Free/trial. Yes, with 10 robotic voices and 1.5x speed.
• Migration. Import PDFs and docs into Speechify for read-aloud, then export audio as MP3 for ad-hoc voiceover work.

Verdict

Pick Speechify Premium when your voice workflow is read-aloud first and creator API second.

7. Amazon Polly — Best for AWS-native apps

Amazon Polly is the right ElevenLabs alternative when you are already paying AWS for compute, storage, or contact center and want predictable per-character TTS pricing. Standard voices cost $4 per 1M characters with a 5M free tier, Neural voices cost $16, Generative voices cost $30, and Long-Form voices cost $100.

• What it does. Synthesizes Standard, Neural, Long-Form, and Generative voices, supports Speech Marks metadata for lipsync and text highlighting, and exposes a fully managed AWS API.
• Best for. AWS-native apps, IVR systems, and teams that already use AWS contact center services.
• Why valid. The Amazon Polly pricing page lists Standard voices at $4 per 1M characters with 5M characters free per month, Neural at $16 with 1M free for 12 months, Long-Form at $100 with 500k free, and Generative at $30 with 100k free for 12 months. New AWS customers also receive up to $200 in AWS Free Tier credits from July 15, 2025.
• Strengths. Predictable per-character billing, AWS IAM, VPC, and GovCloud support, free tier on Standard voices, Generative voice option.
• Limitations. Generative voices feel less expressive than ElevenLabs v3, and Long-Form voices at $100 per 1M characters are the most expensive AWS tier.
• Pricing. Standard $4 per 1M chars; Neural $16/M; Long-Form $100/M; Generative $30/M.
• Free/trial. Yes, 5M Standard characters per month plus 12-month free tiers on Neural, Long-Form, and Generative.
• Migration. Replace ElevenLabs narration with Polly Standard or Neural calls, then upgrade to Generative voices for higher-fidelity audiobooks.

Verdict

Pick Amazon Polly when your stack runs on AWS and you want predictable per-character billing.

8. Resemble AI — Best for voice cloning and AI security

Resemble AI is the strongest ElevenLabs alternative for teams that want voice cloning plus AI security features like watermarking and deepfake detection. The Flex plan is pay-as-you-go with TTS at $0.0005 per second, Rapid voice cloning at $2 per month per voice, and Pro voice cloning at $5 per month per voice, plus a full suite of detection models like DETECT-3B-Omni and Resemblyzer.

• What it does. Generates speech with the Chatterbox and DramaBox research models, supports Rapid and Pro voice cloning, runs voice agents, and bundles Resemble Detect and Resemble Watermarker for AI security.
• Best for. Voice cloning teams, security-conscious enterprises, and contact centers that need deepfake detection.
• Why valid. The Resemble AI pricing page lists the Flex plan with text-to-speech at $0.0005 per second, voice agents at $0.001 per second, Rapid voice clone at $2 per month per voice, Pro voice clone at $5 per month per voice, Team seats at $20 per month per user, and Enterprise with volume discounts up to 80% and on-premise deployment.
• Strengths. Voice cloning à la carte on the Flex plan, watermarking and deepfake detection bundled, credits never expire, on-premise option on Enterprise.
• Limitations. No monthly subscription tier, and per-second usage rates plus clone fees can add up for high-volume creators.
• Pricing. Flex pay-as-you-go; TTS $0.0005/sec; Rapid clone $2/mo; Pro clone $5/mo; Team seats $20/mo; Enterprise custom.
• Free/trial. Yes, $0 to start on the Flex plan with credits never expiring.
• Migration. Re-export ElevenLabs voice clone samples, recreate them on Resemble Rapid or Pro cloning, then turn on Watermarker for production deployments.

Verdict

Pick Resemble AI Flex when voice cloning and AI security are equally important.

9. Murf AI — Best for enterprise content production

Murf AI is the strongest ElevenLabs alternative when enterprise content production matters more than voice cloning or voice-agent workflows. The Murf Falcon API delivers sub-55 ms latency at $0.01 per minute and handles 10,000 concurrent calls, while Murf Gen 2 covers narration with 200+ voices across 35+ languages.

• What it does. Generates narration with Murf Gen 2, runs voice agents with Murf Falcon, supports voice cloning, and ships an AI voice studio with word-level pitch, speed, and emphasis control.
• Best for. Enterprise content teams, learning and development, IVR systems, and dubbing.
• Why valid. The Murf text-to-speech page documents Murf Falcon at $0.01 per minute with sub-55 ms latency and 10,000 concurrent calls, Murf Gen 2 with 99.38% pronunciation accuracy across 200+ voices and 35+ languages, and an AI Voice Studio with pitch, speed, and emphasis controls per word.
• Strengths. Sub-55 ms Falcon latency, 200+ voices across 35+ languages, word-level prosody control, SOC 2 and GDPR posture.
• Limitations. Studio subscription pricing is sales-led rather than published on a single page, and the curated voice library is smaller than ElevenLabs’ community marketplace.
• Pricing. Murf Falcon API $0.01/minute; Gen 2 API per usage; enterprise custom.
• Free/trial. Yes, free online TTS editor with 200+ voices and no sign-up required.
• Migration. Re-author ElevenLabs scripts in Murf Studio, adjust word-level pitch and emphasis, then export finished audio or wire to the Falcon API for production apps.

Verdict

Pick Murf AI when enterprise content production is the primary job and Falcon latency matters for live IVR.

10. Kokoro TTS — Best open-source option

Kokoro-82M is the best open-source ElevenLabs alternative for teams that want to self-host and avoid per-character fees. The model is Apache-2.0 licensed, uses StyleTTS 2 plus iSTFTNet, and was trained on permissive audio and IPA phoneme labels. The hosted API market rate is under $1 per 1M characters of input as of April 2025.

• What it does. Generates speech from text with multiple voice blends and language packs, supports fine-tuning workflows, and runs anywhere you can deploy a Python pipeline.
• Best for. Developers who want to avoid per-character fees, teams that need full control over data and hosting, and hobbyists building privacy-first voice apps.
• Why valid. The Kokoro-82M model card on Hugging Face documents Apache-2.0 weights, 82M parameters, StyleTTS 2 + iSTFTNet architecture, multiple voice blends, and an API market rate under $1 per million characters of input (Replicate at 65¢/M chars, DeepInfra at 80¢/M chars per ArtificialAnalysis, April 2025).
• Strengths. Apache-2.0 license, runs on a single A100, supports multiple languages and voice blends, no per-character fees when self-hosted.
• Limitations. Voice library and ecosystem are tiny compared to ElevenLabs, and you operate the hosting, scaling, and safety guardrails yourself.
• Pricing. Free self-hosted; hosted API under $1 per 1M characters.
• Free/trial. Yes, Apache-2.0 weights are free to download.
• Migration. Port ElevenLabs narration scripts as plain text, run them through the Kokoro pipeline, and host the model on your own infrastructure.

Verdict

Pick Kokoro if you want to self-host and avoid per-character fees at scale.

Which ElevenLabs alternative should you choose?

1. Choose Cartesia Sonic Pro at $5/month if you want the closest drop-in ElevenLabs replacement with voice cloning and real-time agents.
2. Choose Hume Octave Creator at $14/month if empathy and reasoning about tone beat raw expressiveness.
3. Choose OpenAI gpt-realtime if your stack already runs on GPT-5 and you want voice agents in the same billing plan.
4. Choose Google Cloud TTS if you need enterprise scale and Vertex AI integration with predictable multi-region deployment.
5. Choose Azure Speech in Foundry Tools if your team lives inside Microsoft 365 and needs Voice Live API and BAAs for HIPAA.
6. Choose Speechify Premium at $29/month if your voice workflow is read-aloud and dictation first.
7. Choose Amazon Polly if you are already paying AWS and want predictable per-character billing on Standard or Generative voices.
8. Choose Resemble AI Flex if voice cloning plus deepfake detection and watermarking matter more than a polished Studio UI.
9. Choose Murf AI when enterprise content production is the primary job and Falcon latency matters for live IVR.
10. Choose Kokoro if you want to self-host and avoid per-character fees at scale.

When you should keep ElevenLabs

Stay on ElevenLabs when Eleven v3 emotional range, Professional Voice Cloning, Eleven Music v2, or Dubbing v2 are central to your workflow. Eleven v3 went generally available on February 2, 2026 with a 72% user preference over the Alpha release and a 68% reduction in errors on numeric and symbolic pronunciation across 27 categories in 8 languages. Eleven Music v2 launched May 26, 2026 with section-by-section composition, inpainting, and meaningful improvements in multilingual lyrics. Dubbing v2 released May 28, 2026 and preserves emotion and performance across languages for the first time. Pro Voice Cloning is available on the Creator tier and above, and the Iconic Marketplace ships licensed celebrity voices like Sir Michael Caine and Matthew McConaughey.

When switching makes sense

Switch when the credit-pool economics stop working for you, when your voice workflow needs to live inside an existing stack like GPT-5, Vertex AI, or Microsoft 365, or when you want predictable per-character billing instead of credit allotments. Switch when self-hosting and Apache-2.0 weights are a hard requirement, when you need BAAs for HIPAA at a lower price point, or when the model you actually want runs on Cartesia Sonic, Hume Octave, or OpenAI Realtime. Switch when the extras you use (voice cloning, music, dubbing) are cheaper as standalone products on a competitor.

How to migrate without chaos

1. Export every ElevenLabs voice clone sample as WAV or MP3 before you cancel.
2. Re-upload those samples into Cartesia Sonic instant voice cloning, Hume Octave unlimited cloning, or Resemble AI Rapid cloning so the new tool has the same voice identity.
3. Save your ElevenLabs pronunciation dictionaries, voice presets, and Studio project metadata as text or JSON so you can recreate them in the new platform.
4. Run both subscriptions side by side for one billing cycle on the same script to validate voice quality, latency, and cost before cancelling Pro or Scale.
5. Confirm whether your compliance requirements (HIPAA, GDPR, SOC 2, EU data residency) are met by the alternative before you migrate regulated workloads.

Final recommendation

For most readers in 2026, the best ElevenLabs alternative is Cartesia Sonic Pro at $5/month if you want the closest clone of ElevenLabs’ Studio workflow at a fraction of the cost. Pick Hume Octave Creator at $14/month if empathy and voice design matter more than raw expressiveness. Pick OpenAI Realtime if your voice workflow already runs on GPT-5. Pick Google Cloud TTS or Azure Speech for enterprise scale, Amazon Polly for predictable AWS-native billing, Resemble AI for cloning plus security, Murf AI for enterprise content, Speechify for personal reading, and Kokoro for self-hosted inference.

Keep ElevenLabs when Eleven v3 emotional range, Eleven Music v2, Professional Voice Cloning, or Dubbing v2 are non-negotiable. For everything else, at least one alternative on this list will give you more characters per dollar.