Adds WhisperX, reworks UI (still needs some work), theoretically usable

This commit is contained in:
Guillaume Tâche
2024-08-17 22:05:04 +02:00
parent 7bddf53bab
commit 3fa51eb95b
204 changed files with 4787 additions and 1321 deletions

19
whisper/base/pom.xml Normal file
View File

@@ -0,0 +1,19 @@
<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
<modelVersion>4.0.0</modelVersion>
<parent>
<groupId>com.github.gtache.autosubtitle</groupId>
<artifactId>autosubtitle-whisper</artifactId>
<version>1.0-SNAPSHOT</version>
</parent>
<artifactId>autosubtitle-whisper-base</artifactId>
<dependencies>
<dependency>
<groupId>com.github.gtache.autosubtitle</groupId>
<artifactId>autosubtitle-whisper-common</artifactId>
</dependency>
</dependencies>
</project>

View File

@@ -0,0 +1,43 @@
package com.github.gtache.autosubtitle.modules.setup.whisper.base;
import com.github.gtache.autosubtitle.modules.setup.impl.SubtitleExtractorSetup;
import com.github.gtache.autosubtitle.modules.setup.impl.ToolsRoot;
import com.github.gtache.autosubtitle.modules.setup.whisper.PythonVersion;
import com.github.gtache.autosubtitle.modules.setup.whisper.WhisperBundledRoot;
import com.github.gtache.autosubtitle.modules.setup.whisper.WhisperCommonSetupModule;
import com.github.gtache.autosubtitle.modules.setup.whisper.WhisperVenvPath;
import com.github.gtache.autosubtitle.setup.SetupManager;
import com.github.gtache.autosubtitle.setup.whisper.base.WhisperSetupManager;
import dagger.Binds;
import dagger.Module;
import dagger.Provides;
import java.nio.file.Path;
/**
* Setup module for Whisper
*/
@Module(includes = WhisperCommonSetupModule.class)
public abstract class WhisperSetupModule {
@Binds
@SubtitleExtractorSetup
abstract SetupManager bindsSubtitleExtractorSetupManager(final WhisperSetupManager manager);
@Provides
@PythonVersion
static String providesPythonVersion() {
return "3.9.19";
}
@Provides
@WhisperBundledRoot
static Path providesWhisperBundledRoot(@ToolsRoot final Path root) {
return root.resolve("whisper");
}
@Provides
@WhisperVenvPath
static Path providesWhisperVenvPath(@WhisperBundledRoot final Path root) {
return root.resolve("whisper-env");
}
}

View File

@@ -0,0 +1,20 @@
package com.github.gtache.autosubtitle.modules.subtitle.extractor.whisper.base;
import com.github.gtache.autosubtitle.subtitle.extractor.SubtitleExtractor;
import com.github.gtache.autosubtitle.subtitle.extractor.whisper.base.WhisperSubtitleExtractor;
import dagger.Binds;
import dagger.Module;
/**
* Dagger module for Whisper
*/
@Module
public abstract class WhisperExtractorModule {
private WhisperExtractorModule() {
}
@Binds
abstract SubtitleExtractor bindsSubtitleExtractor(final WhisperSubtitleExtractor extractor);
}

View File

@@ -0,0 +1,31 @@
package com.github.gtache.autosubtitle.modules.subtitle.parser.json.whisper.base;
import com.github.gtache.autosubtitle.subtitle.converter.SubtitleConverter;
import com.github.gtache.autosubtitle.subtitle.parser.json.whisper.base.JSONSubtitleConverter;
import com.google.gson.Gson;
import dagger.Binds;
import dagger.Module;
import dagger.Provides;
import dagger.multibindings.IntoMap;
import dagger.multibindings.StringKey;
import javax.inject.Singleton;
@Module
public abstract class WhisperJsonModule {
private WhisperJsonModule() {
}
@Binds
@IntoMap
@StringKey("json")
abstract SubtitleConverter bindsJSONSubtitleConverter(final JSONSubtitleConverter subtitleConverter);
@Provides
@Singleton
static Gson providesGson() {
return new Gson();
}
}

View File

@@ -0,0 +1,23 @@
package com.github.gtache.autosubtitle.modules.whisper.base;
import com.github.gtache.autosubtitle.modules.setup.whisper.base.WhisperSetupModule;
import com.github.gtache.autosubtitle.modules.subtitle.extractor.whisper.base.WhisperExtractorModule;
import com.github.gtache.autosubtitle.modules.subtitle.parser.json.whisper.base.WhisperJsonModule;
import com.github.gtache.autosubtitle.subtitle.extractor.ExtractionModelProvider;
import com.github.gtache.autosubtitle.whisper.WhisperExtractionModelProvider;
import dagger.Binds;
import dagger.Module;
/**
* Dagger module for Whisper
*/
@Module(includes = {WhisperSetupModule.class, WhisperJsonModule.class, WhisperExtractorModule.class})
public abstract class WhisperModule {
private WhisperModule() {
}
@Binds
abstract ExtractionModelProvider bindsExtractionModelProvider(final WhisperExtractionModelProvider provider);
}

View File

@@ -0,0 +1,65 @@
package com.github.gtache.autosubtitle.setup.whisper.base;
import com.github.gtache.autosubtitle.setup.SetupException;
import com.github.gtache.autosubtitle.setup.conda.CondaSetupManager;
import com.github.gtache.autosubtitle.setup.whisper.AbstractWhisperSetupManager;
import com.github.gtache.autosubtitle.setup.whisper.WhisperSetupConfiguration;
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import javax.inject.Inject;
import javax.inject.Singleton;
import java.io.IOException;
import java.nio.file.Files;
import java.time.Duration;
import java.util.List;
/**
* {@link com.github.gtache.autosubtitle.setup.SetupManager} for Whisper
*/
@Singleton
public class WhisperSetupManager extends AbstractWhisperSetupManager {
private static final Logger logger = LogManager.getLogger(WhisperSetupManager.class);
@Inject
WhisperSetupManager(final CondaSetupManager condaSetupManager, final WhisperSetupConfiguration configuration) {
super(condaSetupManager, configuration);
}
@Override
public String name() {
return "Whisper";
}
@Override
protected void installWhisper() throws SetupException {
final var path = getPythonPath();
try {
logger.info("Installing whisper");
final var result = run(List.of(path.toString(), "-m", "pip", "install", "-U", "openai-whisper", "numpy<2"), Duration.ofMinutes(15));
if (result.exitCode() == 0) {
logger.info("Whisper installed");
} else {
throw new SetupException("Error installing whisper: " + result.output());
}
} catch (final IOException e) {
throw new SetupException(e);
}
}
@Override
protected boolean isWhisperInstalled() throws SetupException {
final var path = getPythonPath();
if (Files.exists(path)) {
try {
final var result = run(List.of(path.toString(), "-m", "pip", "show", "openai-whisper"), Duration.ofSeconds(5));
return result.exitCode() == 0;
} catch (final IOException e) {
throw new SetupException(e);
}
} else {
return false;
}
}
}

View File

@@ -0,0 +1,58 @@
package com.github.gtache.autosubtitle.subtitle.extractor.whisper.base;
import com.github.gtache.autosubtitle.Language;
import com.github.gtache.autosubtitle.impl.OS;
import com.github.gtache.autosubtitle.modules.setup.whisper.WhisperVenvPath;
import com.github.gtache.autosubtitle.subtitle.converter.SubtitleConverter;
import com.github.gtache.autosubtitle.subtitle.extractor.ExtractionModel;
import com.github.gtache.autosubtitle.subtitle.extractor.SubtitleExtractor;
import com.github.gtache.autosubtitle.subtitle.extractor.whisper.AbstractWhisperSubtitleExtractor;
import com.github.gtache.autosubtitle.whisper.WhisperModels;
import javax.inject.Inject;
import javax.inject.Singleton;
import java.nio.file.Path;
import java.util.ArrayList;
import java.util.List;
import java.util.Map;
/**
* Whisper implementation of {@link SubtitleExtractor}
*/
@Singleton
public class WhisperSubtitleExtractor extends AbstractWhisperSubtitleExtractor {
@Inject
WhisperSubtitleExtractor(@WhisperVenvPath final Path venvPath, final Map<String, SubtitleConverter> converters, final OS os) {
super(venvPath, converters, os);
}
@Override
protected List<String> createArgs(final Path path, final Language language, final ExtractionModel model, final Path outputDir) {
final var args = new ArrayList<String>(14);
args.add(getPythonPath().toString());
args.add("-m");
args.add("whisper");
args.add("--verbose");
args.add("False");
args.add("--model");
if (model != WhisperModels.LARGE && language == Language.EN) {
args.add(model.name().toLowerCase() + ".en");
} else {
args.add(model.name().toLowerCase());
}
args.add("--task");
args.add("transcribe");
args.add("--output_dir");
args.add(outputDir.toString());
args.add("--output_format");
args.add("json");
if (language != Language.AUTO) {
args.add("--language");
args.add(language.iso2());
}
args.add(path.toString());
return args;
}
}

View File

@@ -0,0 +1,75 @@
package com.github.gtache.autosubtitle.subtitle.parser.json.whisper.base;
import com.github.gtache.autosubtitle.Language;
import com.github.gtache.autosubtitle.subtitle.Subtitle;
import com.github.gtache.autosubtitle.subtitle.SubtitleCollection;
import com.github.gtache.autosubtitle.subtitle.converter.ParseException;
import com.github.gtache.autosubtitle.subtitle.converter.SubtitleConverter;
import com.github.gtache.autosubtitle.subtitle.impl.SubtitleCollectionImpl;
import com.github.gtache.autosubtitle.subtitle.impl.SubtitleImpl;
import com.google.gson.Gson;
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import javax.inject.Inject;
import javax.inject.Singleton;
import java.nio.file.Path;
import java.util.Comparator;
import java.util.List;
import java.util.Objects;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.stream.Collectors;
/**
* {@link SubtitleConverter} implementation for JSON files
*/
@Singleton
public class JSONSubtitleConverter implements SubtitleConverter<SubtitleImpl> {
private static final Logger logger = LogManager.getLogger(JSONSubtitleConverter.class);
private final Gson gson;
@Inject
JSONSubtitleConverter(final Gson gson) {
this.gson = Objects.requireNonNull(gson);
}
@Override
public String format(final SubtitleCollection<?> collection) {
final var id = new AtomicInteger(0);
final var segments = collection.subtitles().stream().map(s -> new JSONSubtitleSegment(id.incrementAndGet(), 0, s.start() / (double) 1000,
s.end() / (double) 1000, s.content(), List.of(), 0, 0, 0, 0)).toList();
final var subtitles = new JSONSubtitles(collection.text(), segments, collection.language().iso2());
return gson.toJson(subtitles);
}
@Override
public SubtitleCollectionImpl<SubtitleImpl> parse(final String content) throws ParseException {
try {
final var json = gson.fromJson(content, JSONSubtitles.class);
final var subtitles = json.segments().stream().map(s -> {
final var start = (long) s.start() * 1000L;
final var end = (long) s.end() * 1000L;
return new SubtitleImpl(s.text(), start, end, null, null);
}).sorted(Comparator.comparing(Subtitle::start).thenComparing(Subtitle::end)).toList();
final var language = Language.getLanguage(json.language());
final var subtitlesText = subtitles.stream().map(Subtitle::content).collect(Collectors.joining(""));
if (!Objects.equals(json.text(), subtitlesText)) {
logger.warn("Not same text: {}\n\n{}", json.text(), subtitlesText);
}
return new SubtitleCollectionImpl<>(json.text(), subtitles, language);
} catch (final Exception e) {
throw new ParseException(e);
}
}
@Override
public boolean canParse(final Path file) {
return file.getFileName().toString().endsWith(".json");
}
@Override
public String formatName() {
return "json";
}
}

View File

@@ -0,0 +1,8 @@
package com.github.gtache.autosubtitle.subtitle.parser.json.whisper.base;
import java.util.List;
public record JSONSubtitleSegment(int id, int seek, double start, double end, String text, List<Integer> tokens,
double temperature, double avg_logprob, double compression_ratio,
double no_speech_prob) {
}

View File

@@ -0,0 +1,6 @@
package com.github.gtache.autosubtitle.subtitle.parser.json.whisper.base;
import java.util.List;
public record JSONSubtitles(String text, List<JSONSubtitleSegment> segments, String language) {
}

View File

@@ -0,0 +1,15 @@
/**
* Base whisper implementation (openai-whisper)
*/
module com.github.gtache.autosubtitle.whisper.base {
requires transitive com.github.gtache.autosubtitle.whisper.common;
requires com.github.gtache.autosubtitle.core;
requires org.apache.logging.log4j;
requires com.google.gson;
exports com.github.gtache.autosubtitle.setup.whisper.base;
exports com.github.gtache.autosubtitle.modules.whisper.base;
exports com.github.gtache.autosubtitle.modules.setup.whisper.base;
exports com.github.gtache.autosubtitle.subtitle.extractor.whisper.base;
exports com.github.gtache.autosubtitle.modules.subtitle.extractor.whisper.base;
}