Generative AI startup Writer Inc. introduced a new voice capability to its platform Tuesday.
The voice capability comes as the generative AI tool continues to grow in the marketing, communications and advertising domain.
Voice allows users to upload a sample copy of a piece of writing, and the Writer platform will analyze the sample and build a voice or style profile. The profile is then applied to any output users want to generate for different purposes.
Writer’s Voice is powered by two of the platform’s language models: voice extraction and voice generation.
Both are part of the Writer’s family of in-house LLM called Palmyra. Palmyra LLMs are trained with business content and marketing data.
Voice extraction is a language model trained to synthesize text examples and build unique voice profiles, according to Writer, based in San Francisco.
Meanwhile, voice generation is a language model trained to generate text that matches the style of a specific voice profile.
Personalizing GenAI tools
The new capability comes as more enterprises favor the personalization of GenAI tools and AI assistants, compared with general-purpose chatbots and assistants.
Writer’s competitor Grammarly also introduced a new feature for personalized writing Tuesday. Strategic suggestions will analyze text and provide feedback to target the text to the right audience.
“What we’re now starting to see is a little bit more personalization actually coming into this GenAI world, where these models — because of their access to the data sources that we have — are able to craft responses that are much more personalized,” said Gartner analyst Arun Chandrasekaran.
Lisa GatelyAnalyst, Forrester
For marketers, a capability like Voice allows them not to leave the brand voice to just prompting, Forrester analyst Lisa Gately said.
“From the time that people have started using generative AI, if you are depending on your users, your employees to do the prompting, you’re leaving that in really subjective hands,” Gately said. However, with Writer’s voice, marketers and experts can ensure that what is generated is on-brand with a company’s message.
However, marketers must also define their brand voice.
“You’ve got to have your act together,” Gately said. “You should have your messaging. You should have some of these artifacts so that in the training … you are confident in how your brand voice should sound.”
Trust and data privacy
Users must also be careful in trusting a tool like this, Chandrasekaran said.
Too much trust could lead to users not cross-checking the facts and information and become disastrous.
“We have to be careful in terms of how much we trust these systems,” Chandrasekaran said. “We still need to have the guardrails and controls in place in order to make sure that we are using these tools in a responsible way.”
There is also the challenge of trusting a vendor like Writer with enterprise data, Chandrasekaran added.
“How much should we trust the providers to ensure that they respect our intellectual property and they respect the privacy of our data,” he continued.
The voice capability has guardrails to ensure safe content, according to Writer.
“Our content is always checked,” said Doris Jwo, director of product management, in an interview. She added that Writer looks at user input to check for toxicity and compliance and ensures that generated outputs are not toxic or aggressive.
The new voice capability will be part of Writer’s current pricing system, Jwo said.
Currently, a team of five can pay up to $18 per month. Writer has custom pricing for enterprise customers.
Esther Ajao is a TechTarget Editorial news writer and podcast host covering artificial intelligence software and systems.