# 144 How André Bastié Is Scaling Happy Scribe After Finding Instant Product-Market Fit

Introduction and Background

In this section, André Bastié introduces himself and discusses the origins of Happy Scribe.

André's Background

• André is the co-founder and CEO of Happy Scribe, based in Barcelona, Spain for three years.

• He studied business-related topics in Paris and completed a Masters in Ecommerce in Ireland.

• The idea for Happy Scribe emerged during his qualitative research on social entrepreneurs in Ireland.

Early Challenges and Inception of Happy Scribe

This part delves into the initial challenges faced by André that led to the creation of Happy Scribe.

Inception Story

• André struggled with transcribing interviews during his research project.

• Collaborated with his flatmate, Marc, who later became the CTO of Happy Scribe.

• Utilized Google Speech-to-Text API to save time transcribing interviews.

Evolution of Happy Scribe's Technology

This segment highlights the technological evolution and early adoption phase of Happy Scribe.

Technological Advancements

• Initial version developed over a weekend with a web interface.

• Positive reception from users post-launch through forums.

• Recognition from a US journalist led to viral growth with 50,000 website views.

Pivotal Moment and Full-Time Commitment

The pivotal moment when André decided to commit full-time to developing Happy Scribe.

Decision Making

• Overwhelmed by viral response prompting full-time dedication to Happy Scribe.

• Transition from startup inception to rapid growth within weeks.

Learning Curve and Industry Insight

Insights gained during the intense early stage of building Happy Scribe.

Industry Understanding

• Initial phase focused on saving time for journalists rather than just transcription tools.

Learning and Evolving with Happy Scribe

In this section, the speaker discusses the evolution of Happy Scribe from an opportunistic approach to a more focused vision on product development and speech-to-text technology.

Evolution of Focus

• Happy Scribe initially adopted an opportunistic approach, saying yes to various opportunities and learning along the way.

• The company shifted focus towards developing the product, particularly enhancing speech-to-text technology by addressing fundamental AI model issues.

Building a Unified Language Platform

This part delves into Happy Scribe's strategy of creating an all-in-one platform for language services to cater to diverse needs efficiently.

Unified Language Platform

• Happy Scribe aimed at building a comprehensive platform for language requirements, integrating transcription, translation, and subtitle services.

• The company strived to offer a one-stop solution for language services catering not only to individuals but also expanding its focus towards SMEs and enterprises.

Transitioning Towards Team Collaboration

Here, the discussion revolves around Happy Scribe's transition from individual customers to focusing more on companies and teams for collaborative language solutions.

Team Collaboration Focus

• Shifting focus towards companies and teams, aiming to facilitate collaboration around languages within organizations.

• Transitioning pricing models from transactional to per seat pricing for team usage, emphasizing collaboration in service offerings.

Funding Strategy and Growth

This segment explores Happy Scribe's funding history, growth strategies, and potential shifts in pricing models towards team-oriented structures.

Funding Approach

• Historically bootstrapped since inception without external investments or significant funding rounds.

• Bootstrapping allowed a focused approach on revenue generation while highlighting the importance of resource optimization in decision-making processes.

Enhancing Speech-to-Text Technology

The conversation centers on the evolution of speech-to-text technology at Happy Scribe through customizations beyond Google Cloud API integration.

Technological Advancements

• Initial reliance on Google Cloud Speech API evolving into multilingual benchmarking across various providers for optimal performance.

New Section

In this section, the speaker discusses the customization options available for customers or subscribers using their platform.

Customization Options

• Customers can add their own vocabularies to improve recognition accuracy.

• The platform allows users to input specific jargon, company names, and acronyms for better transcription results.

• Human-made transcription services are provided, enabling users to build customer dictionaries for consistency.

• Users can customize subtitling aspects such as characters per second and line spacing on a project basis.

• Future plans include storing user voice parameters to save time on repetitive tasks.

New Section

This part delves into the metrics used in transcription quality assessment and the shift towards more subjective evaluation methods.

Transcription Quality Metrics

• Word error rate was commonly used but is now considered less accurate due to advancements in AI technology.

• The focus has shifted towards subjective metrics like word correction rate to capture audio difficulty and subjectivity.

• Internal transcribers' time spent correcting transcripts is monitored for quality evaluation.

New Section

The discussion centers around machine translation features within the platform and its importance compared to transcription services.

Machine Translation Features

• The platform collaborates with DeepL and Google Translate for machine translation services.

• DeepL handles languages not supported by Google Translate.

• Current focus remains on foreign subtitles rather than extensive translation services.

New Section

This segment explores the increasing demand for subtitling solutions driven by the rise of video content consumption without sound.

Subtitling Demand

• Subtitles play a crucial role in capturing viewer attention when audio is off by default on many platforms.

New Section

In this section, the speaker discusses interesting metrics related to video consumption habits.

Video Consumption Metrics

• Subtitles without audio:

• Around 92% of people watch videos with subtitles and no audio.

• Sound-off in public spaces:

• 69% of individuals watch videos with sound off in public spaces, emphasizing the importance of subtitles for silent viewing.

New Section

The conversation shifts towards influencer agencies' growing interest in subtitling content creators' videos.

Influencer Agencies and Subtitling

• Growing interest from influencer agencies:

• Influencer agencies are increasingly utilizing subtitling services for their content creators, highlighting a shift towards outsourcing this task.

• Importance of accessibility:

• Subtitling impacts accessibility, catering to diverse audience needs based on age groups and content preferences.

New Section

The discussion delves into the nuances of subtitling across different video formats and target audiences.

Nuances of Subtitling

• Varied character limits per line:

• Different video formats require specific character limits per line, influenced by factors such as audience age and reading speed.

• Artistic nature of subtitling:

• Subtitling is described as more art than science, requiring deep language understanding to ensure readability tailored to the target audience.

New Section

The focus shifts to human-made translation services offered by Happy Scribe and its impact on business model evolution.

Human-Made Translation Services

• Evolution into human-made services:

• Happy Scribe introduces human-made translation services based on user demand and market trends, enhancing customer experience and expanding service offerings.

• Balancing automation and human touch:

• Combining AI transcription with human proofreading provides value for customers handling large volumes of transcription work while maintaining quality standards.

New Section

This segment explores Happy Scribe's automated processes and standardized guidelines for efficient service delivery.

Automated Processes and Standardization

• Streamlined workflow automation:

• Happy Scribe emphasizes automated processes from file upload to payment, ensuring efficiency while adhering to standardized guidelines for consistent quality output.

• Focus on standard solutions:

Business Model and Service Differentiation

In this section, the speaker discusses the business model of their service, emphasizing standardization, accuracy guarantees, and turnaround times.

Business Model Details

• The business offers a platform where users can request human or AI transcription services with a high level of standardization.

• Guarantees 99% accuracy with job rework or refund policy if not met.

• Ensures 24-hour turnaround time; offers expedited options for quicker transcription.

• Aims to provide predictability and accessibility with a 24/7 service availability.

Enterprise Focus and Product Development

This part delves into the company's focus on enterprise customers, differentiation from competitors like Verbit, and long-term strategic considerations.

Enterprise Strategy Discussion

• Current focus is on teams and SMBs rather than large enterprises.

• Differentiates from competitors by prioritizing product-level enterprise solutions over customized models.

• Long-term strategy leans towards building a fully automated SaaS platform rather than an enterprise-focused approach similar to competitors.

AI Models Evolution and Impact

The conversation shifts towards discussing advancements in AI models like Whisper, their impact on transcription services, and the need for continuous improvement in adapting to evolving events.

AI Model Advancements

• Whisper represents a significant paradigm shift in AI due to its data volume impact and multilingual capabilities.

• Happy Scribe builds upon Whisper's model through fine-tuning for improved performance.

• Challenges include ensuring models adapt to current events seamlessly without breaking during major occurrences like pandemics or conflicts.

Building Unbiased AI Models

Focuses on the importance of developing unbiased AI models that are inclusive of diverse accents and genders while staying updated with current news trends for accurate transcriptions.

Developing Inclusive AI Models

• Emphasis on addressing biases towards men and minority accents in existing AI models.

• Necessity to train models constantly on up-to-date data to handle news-related terms effectively.

Open Source Nature of Whisper Model

Explores the open-source nature of the Whisper model, its accessibility for developers, comparison with other released models like Facebook's speech-to-text model, and implications for future developments.

Open Source Model Insights

• Whisper model is entirely open source, facilitating easier access for developers compared to proprietary systems.

• Mention of other notable open-source releases contributing to advancements in speech recognition technology.

New Section

In this section, the discussion revolves around the potential impact of combining ChatGPT with actual content like transcription and speech-to-text engines.

The Impact of Pairing ChatGPT with Content

• The current hype focuses on irrelevant topics and unimportant questions. However, when paired with actual content such as transcription, the impact can be significant.

• Experimentation with transcribing meetings and summarizing them in bullet points has shown impressive results.

• Envisioning ChatGPT coupled with a flawless speech-to-text engine akin to a functional Alexa is highlighted as a long-awaited solution.

Building Front End Capabilities

• Emphasis is placed on the importance of building a strong front end to complement existing back-end capabilities.

• Happy Scribe's expertise lies in creating user-friendly products for transcription, subtitles, and translation, offering seamless solutions for various users.

Roadmap Insights and Future Launches

• Changes in pricing are anticipated along with ongoing product developments aimed at enhancing user experience.

• Happy Scribe's focus is not solely on transcription or subtitles but on enabling users to leverage these outputs effectively for their intended purposes.