Precisão do Whisper: como melhorar a qualidade da transcrição

• Programadores que otimizam a precisão de transcrição com Whisper
• Criadores que transcrevem podcasts e vídeos
• Investigadores que trabalham com áudio
• Quem procura dicas de precisão do Whisper

O que influencia a precisão do Whisper

• Qualidade do áudio (o mais importante)
• Tamanho do modelo (escolha)
• Deteção de idioma
• Pré-processamento de áudio
• Parâmetros de configuração
• Duração do áudio e segmentação

Dica 1: Escolha o tamanho de modelo certo

import whisper

# Model sizes from fastest to most accurate:
# tiny, base, small, medium, large

# For maximum accuracy, use medium or large
model = whisper.load_model("medium")  # Best balance
# or
model = whisper.load_model("large")  # Maximum accuracy

import whisper

def transcribe_with_optimal_model(audio_path, prioritize_accuracy=True):
    """
    Select model based on accuracy vs speed priority.
    
    Args:
        audio_path: Path to audio file
        prioritize_accuracy: True for accuracy, False for speed
    """
    if prioritize_accuracy:
        model_size = "medium"  # or "large" for best accuracy
    else:
        model_size = "base"  # or "small" for balanced
    
    model = whisper.load_model(model_size)
    result = model.transcribe(audio_path)
    
    return result

# For critical transcriptions
result = transcribe_with_optimal_model("important_meeting.mp3", prioritize_accuracy=True)

Dica 2: Especifique o idioma quando souber

import whisper

model = whisper.load_model("base")

# Auto-detect (less accurate)
result_auto = model.transcribe("audio.mp3")

# Specify language (more accurate)
result_en = model.transcribe("audio.mp3", language="en")
result_zh = model.transcribe("audio.mp3", language="zh")
result_es = model.transcribe("audio.mp3", language="es")

• Menos erros na deteção de idioma
• Melhores resultados para falantes multilingues
• Processamento mais rápido (salta a deteção)
• Melhor tratamento de sotaques e dialetos

import whisper
import langdetect

def transcribe_with_language_detection(audio_path, model_size="base"):
    """
    Detect language first, then transcribe with explicit language.
    """
    model = whisper.load_model(model_size)
    
    # Quick language detection
    result_quick = model.transcribe(audio_path, language=None)
    detected_lang = result_quick["language"]
    
    # Re-transcribe with detected language for better accuracy
    result = model.transcribe(audio_path, language=detected_lang)
    
    return result

result = transcribe_with_language_detection("audio.mp3")

Dica 3: Pré-processe o áudio antes de transcrever

import whisper
import numpy as np
from scipy.io import wavfile
from scipy import signal

def preprocess_audio(audio_path, output_path):
    """
    Preprocess audio to improve transcription accuracy.
    """
    # Read audio file
    sample_rate, audio = wavfile.read(audio_path)
    
    # Normalize audio (scale to [-1, 1])
    if audio.dtype == np.int16:
        audio = audio.astype(np.float32) / 32768.0
    elif audio.dtype == np.int32:
        audio = audio.astype(np.float32) / 2147483648.0
    
    # Remove DC offset
    audio = audio - np.mean(audio)
    
    # Normalize volume
    max_val = np.max(np.abs(audio))
    if max_val > 0:
        audio = audio / max_val * 0.95  # Leave headroom
    
    # Resample to 16kHz (Whisper's optimal sample rate)
    if sample_rate != 16000:
        num_samples = int(len(audio) * 16000 / sample_rate)
        audio = signal.resample(audio, num_samples)
        sample_rate = 16000
    
    # Save preprocessed audio
    wavfile.write(output_path, sample_rate, (audio * 32767).astype(np.int16))
    
    return output_path

# Usage
preprocessed = preprocess_audio("raw_audio.wav", "preprocessed.wav")
model = whisper.load_model("base")
result = model.transcribe(preprocessed)

1. Normalizar níveis – volume uniforme
2. Remover offset DC – eliminar viés constante
3. Reamostrar a 16 kHz – taxa ideal para o Whisper
4. Remover silêncio – focar na fala
5. Reduzir ruído – limpar o fundo

Dica 4: Definições de temperatura para melhores resultados

import whisper

model = whisper.load_model("base")

# Default temperature (0.0)
result_default = model.transcribe("audio.mp3")

# Lower temperature for more deterministic results
result_low_temp = model.transcribe(
    "audio.mp3",
    temperature=0.0,  # Most deterministic
    best_of=5,  # Try multiple decodings, pick best
    beam_size=5  # Beam search size
)

• temperature=0.0: Mais determinístico, melhor precisão
• temperature=0.2: Pouca aleatoriedade, bom equilíbrio
• temperature=0.6: Predefinido, equilibrado
• Valores altos: mais “criativo”, menos preciso

def transcribe_with_optimal_settings(audio_path, model_size="base"):
    """
    Use optimal settings for maximum accuracy.
    """
    model = whisper.load_model(model_size)
    
    result = model.transcribe(
        audio_path,
        temperature=0.0,  # Most deterministic
        best_of=5,  # Try 5 decodings, pick best
        beam_size=5,  # Beam search
        patience=1.0,  # Patience for beam search
        condition_on_previous_text=True,  # Use context
        initial_prompt="This is a conversation about technology."  # Context hint
    )
    
    return result

Dica 5: Prompt inicial para contexto

import whisper

model = whisper.load_model("base")

# Without context
result_basic = model.transcribe("meeting.mp3")

# With context (much better accuracy)
result_context = model.transcribe(
    "meeting.mp3",
    initial_prompt="This is a business meeting discussing project timelines and deliverables."
)

# For technical content
result_tech = model.transcribe(
    "lecture.mp3",
    initial_prompt="This is a computer science lecture about machine learning and neural networks."
)

• Conteúdo técnico: incluir termos do domínio
• Nomes e lugares: mencionar nomes próprios importantes
• Sotaques: descrever sotaque ou dialeto
• Contexto: descrever o cenário ou o tema

def transcribe_with_context(audio_path, context_description):
    """
    Transcribe with context for better accuracy.
    """
    model = whisper.load_model("medium")
    
    result = model.transcribe(
        audio_path,
        initial_prompt=context_description,
        language="en"
    )
    
    return result

# Example usage
result = transcribe_with_context(
    "interview.mp3",
    "This is an interview with Dr. Sarah Johnson about medical research. "
    "The conversation includes technical medical terminology."
)

Dica 6: Trate ficheiros de áudio longos corretamente

import whisper
from pydub import AudioSegment
import os

def transcribe_long_audio(audio_path, model_size="base", chunk_length_minutes=30):
    """
    Transcribe long audio by splitting into optimal chunks.
    """
    model = whisper.load_model(model_size)
    
    # Load audio
    audio = AudioSegment.from_file(audio_path)
    chunk_length_ms = chunk_length_minutes * 60 * 1000
    
    # Split into chunks
    chunks = []
    for i in range(0, len(audio), chunk_length_ms):
        chunks.append(audio[i:i + chunk_length_ms])
    
    # Transcribe each chunk
    full_text = []
    all_segments = []
    
    for i, chunk in enumerate(chunks):
        chunk_path = f"temp_chunk_{i}.wav"
        chunk.export(chunk_path, format="wav")
        
        print(f"Transcribing chunk {i+1}/{len(chunks)}")
        result = model.transcribe(chunk_path)
        
        # Adjust timestamps for chunk offset
        offset = i * chunk_length_ms / 1000.0
        for segment in result["segments"]:
            segment["start"] += offset
            segment["end"] += offset
            all_segments.append(segment)
        
        full_text.append(result["text"])
        
        # Clean up
        os.remove(chunk_path)
    
    # Combine results
    combined_result = {
        "text": " ".join(full_text),
        "segments": all_segments,
        "language": result["language"]
    }
    
    return combined_result

# Usage
result = transcribe_long_audio("long_podcast.mp3", model_size="medium", chunk_length_minutes=30)

• Dividir em blocos de 20–30 minutos
• Usar o mesmo tamanho de modelo em todos os blocos
• Manter contexto entre blocos
• Fundir segmentos com marcas temporais corretas

Dica 7: Otimize para áudio ruidoso

import whisper
import noisereduce as nr
import soundfile as sf
import numpy as np

def transcribe_noisy_audio(audio_path, model_size="medium"):
    """
    Reduce noise before transcription for better accuracy.
    """
    # Load audio
    audio, sample_rate = sf.read(audio_path)
    
    # Reduce noise
    reduced_noise = nr.reduce_noise(
        y=audio,
        sr=sample_rate,
        stationary=False,  # For non-stationary noise
        prop_decrease=0.8  # Reduce noise by 80%
    )
    
    # Save cleaned audio
    cleaned_path = "cleaned_audio.wav"
    sf.write(cleaned_path, reduced_noise, sample_rate)
    
    # Transcribe with larger model (better for noisy audio)
    model = whisper.load_model(model_size)
    result = model.transcribe(cleaned_path)
    
    # Clean up
    os.remove(cleaned_path)
    
    return result

# Usage
result = transcribe_noisy_audio("noisy_recording.mp3", model_size="medium")

• Use modelos medium ou large
• Pré-processe com redução de ruído
• Aumente best_of
• Forneça contexto sobre as condições de ruído

Dica 8: Marcas temporais por palavra para mais controlo

import whisper

model = whisper.load_model("base")

# Get word timestamps
result = model.transcribe(
    "audio.mp3",
    word_timestamps=True  # Enable word-level timestamps
)

# Access word timestamps
for segment in result["segments"]:
    print(f"Segment: {segment['text']}")
    print(f"Start: {segment['start']:.2f}s, End: {segment['end']:.2f}s")
    
    if "words" in segment:
        for word in segment["words"]:
            print(f"  Word: {word['word']} ({word['start']:.2f}s - {word['end']:.2f}s)")

• Legendas: sincronização precisa palavra a palavra
• Correção de erros: localizar palavras problemáticas
• Pesquisa: encontrar palavras na transcrição
• Análise da fala: padrões de fala

Dica 9: Combine várias decodificações

import whisper

model = whisper.load_model("base")

# Single decoding (default)
result_single = model.transcribe("audio.mp3")

# Multiple decodings, pick best (more accurate)
result_best = model.transcribe(
    "audio.mp3",
    best_of=5,  # Try 5 decodings
    temperature=(0.0, 0.2, 0.4, 0.6, 0.8)  # Different temperatures
)

• Precisão: maior com várias decodificações
• Velocidade: mais lenta (5× com best_of=5)
• Quando: precisão crítica, velocidade secundária

Dica 10: Pós-processe as transcrições

import re
import whisper

def post_process_transcript(text):
    """
    Fix common transcription errors.
    """
    # Fix common contractions
    text = re.sub(r"\b(\w+) '(\w+)\b", r"\1'\2", text)  # Fix spacing in contractions
    
    # Fix common homophones (add your own)
    replacements = {
        "there": "their",  # Context-dependent
        "its": "it's",  # Context-dependent
        # Add more based on your domain
    }
    
    # Capitalize sentences
    sentences = re.split(r'([.!?]\s+)', text)
    capitalized = []
    for i, sentence in enumerate(sentences):
        if sentence.strip():
            capitalized.append(sentence[0].upper() + sentence[1:] if len(sentence) > 1 else sentence.upper())
        else:
            capitalized.append(sentence)
    
    return "".join(capitalized)

# Usage
model = whisper.load_model("base")
result = model.transcribe("audio.mp3")
processed_text = post_process_transcript(result["text"])

Exemplo completo: otimização de precisão pronta para produção

import whisper
import os
from pathlib import Path

def transcribe_with_maximum_accuracy(
    audio_path,
    model_size="medium",
    language=None,
    context_prompt=None,
    output_format="txt"
):
    """
    Transcribe audio with maximum accuracy using best practices.
    
    Args:
        audio_path: Path to audio file
        model_size: Whisper model size (medium or large recommended)
        language: Language code (None for auto-detect)
        context_prompt: Initial prompt for context
        output_format: Output format (txt, json, srt)
    """
    # Load model (medium or large for best accuracy)
    print(f"Loading Whisper model: {model_size}")
    model = whisper.load_model(model_size)
    
    # Prepare transcription parameters
    transcribe_kwargs = {
        "temperature": 0.0,  # Most deterministic
        "best_of": 5,  # Try multiple decodings
        "beam_size": 5,  # Beam search
        "patience": 1.0,
        "condition_on_previous_text": True,
        "word_timestamps": True,  # Get word-level timestamps
    }
    
    # Add language if specified
    if language:
        transcribe_kwargs["language"] = language
    
    # Add context prompt if provided
    if context_prompt:
        transcribe_kwargs["initial_prompt"] = context_prompt
    
    # Transcribe
    print(f"Transcribing: {audio_path}")
    result = model.transcribe(audio_path, **transcribe_kwargs)
    
    # Post-process
    result["text"] = post_process_transcript(result["text"])
    
    # Save result
    base_name = Path(audio_path).stem
    output_path = f"{base_name}_transcript.{output_format}"
    
    if output_format == "txt":
        with open(output_path, "w", encoding="utf-8") as f:
            f.write(result["text"])
    elif output_format == "json":
        import json
        with open(output_path, "w", encoding="utf-8") as f:
            json.dump(result, f, indent=2, ensure_ascii=False)
    
    print(f"✓ Transcription saved: {output_path}")
    print(f"  Language: {result['language']}")
    print(f"  Duration: {result['segments'][-1]['end']:.2f}s")
    
    return result

# Example usage
result = transcribe_with_maximum_accuracy(
    audio_path="important_meeting.mp3",
    model_size="medium",
    language="en",
    context_prompt="This is a business meeting discussing quarterly results and project updates.",
    output_format="txt"
)

Comparação de precisão: antes e depois da otimização

Resumo de boas práticas

Para máxima precisão:

1. ✅ Modelo medium ou large
2. ✅ Idioma explícito
3. ✅ Contexto com initial_prompt
4. ✅ temperature=0.0 para resultados determinísticos
5. ✅ word_timestamps para saída detalhada
6. ✅ Pré-processar áudio ruidoso
7. ✅ Dividir ficheiros longos
8. ✅ best_of=5 para conteúdo crítico

Para equilíbrio velocidade/precisão:

1. ✅ Modelo small ou base
2. ✅ Deteção automática de idioma
3. ✅ Temperatura predefinida
4. ✅ Sem best_of
5. ✅ Ficheiros com pré-processamento mínimo

Erros comuns

❌ Modelo tiny para conteúdo importante

❌ Não especificar o idioma

❌ Ignorar o contexto

❌ Predefinições com muito ruído

❌ Ficheiros muito longos de uma vez

Resolução de problemas de precisão

Problema: Baixa precisão em termos técnicos

result = model.transcribe(
    "technical_audio.mp3",
    initial_prompt="This audio contains technical terminology related to machine learning, neural networks, and deep learning."
)

Problema: Má precisão com sotaques

# Use larger model
model = whisper.load_model("medium")

# Provide accent context
result = model.transcribe(
    "accented_audio.mp3",
    initial_prompt="This speaker has a British accent.",
    language="en"
)

Problema: Erros em nomes próprios

# Include names in initial prompt
result = model.transcribe(
    "interview.mp3",
    initial_prompt="This interview features Dr. Sarah Johnson and Professor Michael Chen discussing research."
)

Conclusão

• Modelo: medium ou large para conteúdo crítico
• Configuração: temperatura e decodificação ótimas
• Contexto: informação de domínio
• Pré-processamento: áudio limpo antes de transcrever
• Pós-processamento: corrigir erros comuns automaticamente

1. O tamanho do modelo tem o maior impacto
2. Especificar o idioma melhora muito os resultados
3. Prompts de contexto ajudam em conteúdo especializado
4. Várias decodificações (best_of) aumentam a precisão mas abrandam
5. A qualidade do áudio continua a ser o fator principal