{"id":16448,"date":"2026-11-17T11:04:41","date_gmt":"2026-11-17T02:04:41","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=16448"},"modified":"2026-11-17T11:14:07","modified_gmt":"2026-11-17T02:14:07","slug":"soprano-1-1-80m-gguf-released","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/soprano-1-1-80m-gguf-released\/","title":{"rendered":"Soprano-1.1-80M-GGUF Lightweight English TTS Model: File List"},"content":{"rendered":"<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/huggingface.co\/ggml-org\/Soprano-1.1-80M-GGUF\">ggml-org\/Soprano-1.1-80M-GGUF<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-11-16<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>apache-2.0<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-gguf-en\/\">GGUF<\/a><\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p><code>ggml-org\/Soprano-1.1-80M-GGUF<\/code>, a GGUF format version of the open-weight, lightweight Text-to-Speech (TTS) model &#8220;Soprano-1.1-80M&#8221; that accepts text input and outputs synthesized speech, has been released.<\/p>\n<p>This model is a derivative model based on &#8220;ekwek\/Soprano-1.1-80M&#8221; developed by ekwek, which was converted into the GGUF format using ggml-org&#8217;s automated conversion tool. It is designed to perform fast and expressive speech synthesis on user devices. It is compatible with the llama.cpp ecosystem, making it possible to generate speech directly by calling it from the command-line tool <code>llama-tts<\/code> or the <code>\/tts<\/code> endpoint of the <code>llama serve<\/code> server function.<\/p>\n<h2>Specifications<\/h2>\n<p>Specifications and details confirmed from public documentation and the base model&#8217;s model card are as follows:<\/p>\n<ul>\n<li>Parameters: 80M (approx. 80 million parameters)<\/li>\n<li>Base model architecture: Qwen3ForCausalLM (model type: qwen3)<\/li>\n<li>Input format: Text (normalized text where alphanumeric characters are spelled out beforehand is recommended)<\/li>\n<li>Output format: Audio (wav format, etc.)<\/li>\n<li>Output specifications: Outputs using a single fixed voice. Voice cloning features are not supported<\/li>\n<li>Supported languages: English only<\/li>\n<li>Distribution format: GGUF format (the original model is released in safetensors format)<\/li>\n<li>License terms: Apache-2.0 license (an open-source license permitting commercial use, modification, redistribution, etc.)<\/li>\n<li>Development and conversion: The original model was created by ekwek, and the GGUF version was generated by ggml-org using an automated conversion script.<\/li>\n<\/ul>\n<h2>Performance and Quality<\/h2>\n<p>Information regarding the quality and benchmark figures of this model is published in the model card of the base model, &#8220;ekwek\/Soprano-1.1-80M&#8221;. While benchmark results specific to the GGUF version are not listed, the following performance data and quality characteristics have been reported for the original model.<\/p>\n<p>In comparison with the first-generation version &#8220;Soprano-80M&#8221;, Soprano-1.1-80M is reported to have reduced hallucinations by 95%. Additionally, in user comparisons such as blind tests, it is reported to have achieved a 63% preference rate over the older version.<\/p>\n<p>Regarding generation speed and latency, extremely high processing performance has been published by the developers.<\/p>\n<ul>\n<li>Generation speed in GPU environments: When given inputs of sufficient length, it is reported capable of speech generation at up to 2000x real-time speed<\/li>\n<li>Generation speed in CPU environments: Capable of speech generation at 20x real-time speed<\/li>\n<li>Streaming latency: It is reported capable of streaming audio with a latency of less than 15 milliseconds (&lt;15 ms)<\/li>\n<\/ul>\n<p>On the other hand, the model card also clearly notes quality-related limitations and precautions. The amount of audio data used to train this model is limited to approximately 1,000 hours, which is about 1\/100th of the data scale compared to general large-scale TTS models. Therefore, it is explained that uncommon English words may be mispronounced.<\/p>\n<p>Furthermore, numbers and some special symbols may also be mispronounced, so to obtain the best output, it is recommended to input text in a format where expressions like &#8220;1+1&#8221; are expanded into words such as &#8220;one plus one&#8221;. Regarding sentence length, the most stable quality is reportedly obtained when inputting text so that the generated speech falls within a range of 2 to 15 seconds per sentence.<\/p>\n<h2>Strengths and Use Cases<\/h2>\n<p>&#8220;Soprano-1.1-80M&#8221; is a model specialized in synthesizing English speech extremely quickly and with low latency in on-device environments. Despite having a compact parameter count of 80M (approx. 80 million), the base model has been developed aiming for expressive and high-fidelity speech synthesis.<\/p>\n<p>Specific intended use cases and scenarios where this model excels include:<\/p>\n<ul>\n<li>Real-time voice interaction in local and edge environments: Because playback can begin with an extremely short latency of less than 15 milliseconds during streaming, it is well-suited for response voices of locally running AI assistants or integration into interactive interfaces requiring real-time responsiveness.<\/li>\n<li>Fast English text-to-speech on devices: Thanks to fast inference capabilities of up to 2000x real-time in GPU environments and 20x real-time in CPU environments, it is suited for batch conversion of long texts or background reading processing where device resource consumption needs to be minimized.<\/li>\n<li>Integrated operation with the llama.cpp ecosystem: Through the newly released GGUF version, users can easily utilize speech synthesis capabilities via standalone commands or a lightweight HTTP server through llama.cpp\u2014a lightweight C\/C++ based runtime environment\u2014without relying on external large-scale frameworks.<\/li>\n<\/ul>\n<p>On the other hand, when considering use cases, attention should be paid to the following characteristics and limitations:<\/p>\n<ul>\n<li>The type of voice is limited to a single fixed voice, and it does not support voice cloning functions to reproduce a specific speaker&#8217;s voice.<\/li>\n<li>The supported language is English only. It does not support multilingual synthesis.<\/li>\n<li>To obtain optimal quality, it is recommended to break down texts so that the length per sentence is around 2 to 15 seconds. Also, performing preliminary text normalization\u2014such as writing out numbers and symbols like &#8220;1+1&#8221; as &#8220;one plus one&#8221;\u2014makes it easier to prevent mispronunciations. Avoiding grammatical flaws like not using contractions, and avoiding unnecessary consecutive spaces, is also recommended.<\/li>\n<\/ul>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 80M parameters (taken from the base model ekwek\/Soprano-1.1-80M)<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Your VRAM<\/th>\n<th>Quantization<\/th>\n<th>File size<\/th>\n<th>Est. memory needed<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>4GB (laptop iGPU \/ phone class)<\/td>\n<td>F32<\/td>\n<td>0.3GB<\/td>\n<td>0.4GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model&#8217;s authors. The listed files are distributed in F32 (32-bit). Loading in a lower precision (BF16 or quantized) needs less memory, so treat this as the size of the files as distributed, not a minimum. This site has not measured the lower-precision figures. Compare with other models in our <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a>. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:hardware --><\/p>\n<p><!-- lmw:runnability --><\/p>\n<h2>Can You Run It Locally?<\/h2>\n<p>The publisher distributes this model as GGUF.<\/p>\n<p><strong>License \u2014 <code>apache-2.0<\/code> (Commercial use allowed):<\/strong> Permits commercial use, modification and redistribution. Redistribution requires including the license and stating changes; includes a patent grant.<\/p>\n<p><em>Compiled by this site&#8217;s code from the published formats and the license field. License summaries are not legal advice \u2014 check the publisher&#8217;s original terms before relying on them.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:peers --><\/p>\n<h2>Recent Models in the Same Size Class<\/h2>\n<p><em>Models with <\/em><em>up to 4B<\/em><em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site&#8217;s estimates; licenses are as stated on the model cards.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<p class=\"lmw-table-hint\" style=\"margin:0 0 4px;font-size:0.85em;opacity:0.7;\">\u2192 Scroll horizontally to see all columns<\/p>\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Model<\/th>\n<th>Parameters<\/th>\n<th>Smallest VRAM tier<\/th>\n<th>License<\/th>\n<th>Our article<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>autotrust\/TTS-V3<\/td>\n<td>1.9B<\/td>\n<td>8GB<\/td>\n<td>component-specific-terms<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/autotrust-tts-v3-released\/\">TTS-V3 Phone-Dialogue Specialized TTS Model: 8GB+ VRAM, File List<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>autotrust\/TTS-V4<\/td>\n<td>1.9B<\/td>\n<td>8GB<\/td>\n<td>other<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/autotrust-tts-v4-chinese-call-center-tts\/\">TTS-V4 Chinese Text-to-Speech Model: 8GB+ VRAM, File List<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>phasefield-audio\/Irodori-TTS-v4.1-Anime<\/td>\n<td>766M<\/td>\n<td>4GB<\/td>\n<td>mit<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/07\/irodori-tts-v41-anime-released\/\">Irodori-TTS-v4.1-Anime: Our Generated Audio, File List<\/a> (2026-09-07)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:peers --><\/p>\n<p><!-- lmw:files --><\/p>\n<h2>Distributed Files<\/h2>\n<p><em>Weight files published in <a href=\"https:\/\/huggingface.co\/ggml-org\/Soprano-1.1-80M-GGUF\/tree\/main\">ggml-org\/Soprano-1.1-80M-GGUF<\/a>, listed by this site from the Hugging Face API. Sizes are the actual file sizes.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>File<\/th>\n<th>Size<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><code>mmproj-Soprano-1.1-80M-F16.gguf<\/code><\/td>\n<td>61MB<\/td>\n<\/tr>\n<tr>\n<td><code>Soprano-1.1-80M-F32.gguf<\/code><\/td>\n<td>319MB<\/td>\n<\/tr>\n<tr>\n<td><code>Soprano-1.1-80M-Q8_0.gguf<\/code><\/td>\n<td>85MB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:files --><\/p>\n<h2>How to Get It<\/h2>\n<p>This model is distributed in GGUF format and can be obtained from the Hugging Face repository. Since it is not a gated model, it can be downloaded without prior application or consent procedures.<\/p>\n<ul>\n<li>Repository: <a href=\"https:\/\/huggingface.co\/ggml-org\/Soprano-1.1-80M-GGUF\">ggml-org\/Soprano-1.1-80M-GGUF<\/a><\/li>\n<li>Distribution format: GGUF<\/li>\n<\/ul>\n<p>By utilizing the llama.cpp toolset, direct execution from the terminal and startup as a server are possible. The execution steps listed in the model card are as follows:<\/p>\n<h3>Direct Execution from Terminal<\/h3>\n<p>By using the <code>llama-tts<\/code> command, you can directly specify the Hugging Face repository to generate an audio file.<\/p>\n<pre><code class=\"language-bash\">llama-tts -hf ggml-org\/Soprano-1.1-80M-GGUF -p &quot;Hello world&quot; --output out.wav\n<\/code><\/pre>\n<h3>Starting as a Server<\/h3>\n<p>You can launch a server using the <code>llama serve<\/code> command and call speech synthesis via the <code>\/tts<\/code> endpoint.<\/p>\n<pre><code class=\"language-bash\">llama serve -hf ggml-org\/Soprano-1.1-80M-GGUF\n<\/code><\/pre>\n<p>This model was automatically converted using <a href=\"https:\/\/github.com\/ggml-org\/convert\">ggml-org\/convert<\/a>. Detailed documentation on TTS features in llama.cpp and the latest usage methods can be checked at <a href=\"https:\/\/github.com\/ggml-org\/llama.cpp\/blob\/master\/docs\/multimodal\/tts.md#soprano\">llama.cpp TTS documentation<\/a>.<\/p>\n<p>Additionally, the original conversion source model is published in the following repository:<\/p>\n<ul>\n<li>Base model: <a href=\"https:\/\/huggingface.co\/ekwek\/Soprano-1.1-80M\">ekwek\/Soprano-1.1-80M<\/a><\/li>\n<\/ul>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/higgs-audio-v3-stt-gguf\/\">higgs-audio-v3-stt-GGUF Speech-to-Text Model: 4GB+ VRAM, File List<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/run-resemble-ai-chatterbox-turbo-locally-gguf\/\">chatterbox-turbo-GGUF Text-to-Speech Model: 4GB+ VRAM, File List<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/chatterbox-multilingual-v3-gguf-2\/\">chatterbox-GGUF 23-Language Multilingual TTS Model: 4GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/voxtral-mini-4b-realtime-2602-gguf-2\/\">Voxtral-Mini-4B-Realtime-2602-GGUF Real-time ASR Model: 4GB+ VRAM<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Find models by VRAM<\/strong> (This model runs from the 4GB tier) \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-8gb-en\/\">Other models that run on a 8GB GPU<\/a><\/li>\n<li><strong>Formats this model is available in<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-gguf-en\/\">GGUF format guide and models<\/a><\/li>\n<li><strong>Other models for the same task<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/models-by-task-en\/#task-audio\">Other audio models<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/ggml-org\/Soprano-1.1-80M-GGUF\">ggml-org\/Soprano-1.1-80M-GGUF<\/a><\/li>\n<li><a href=\"https:\/\/huggingface.co\/ekwek\/Soprano-1.1-80M\">ekwek\/Soprano-1.1-80M<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/ggml-org\/llama.cpp\/blob\/master\/docs\/multimodal\/tts.md#soprano\">llama.cpp TTS documentation<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/ggml-org\/convert\">ggml-org\/convert<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Explore Soprano-1.1-80M-GGUF, an open-weight 80M Text-to-Speech model in GGUF format for fast, low-latency English speech synthesis on local hardware.<\/p>\n","protected":false},"author":1,"featured_media":16447,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[203],"tags":[161,163,1073,5076,5078,1547,213],"class_list":["post-16448","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-image-video-and-audio","tag-ggml-en","tag-gguf-en","tag-llama-cpp-en","tag-soprano-1-1-en","tag-soprano-1-1-80m-en","tag-verified","tag--en"],"lang":"en","translations":{"en":16448,"ja":16446},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16448","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=16448"}],"version-history":[{"count":2,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16448\/revisions"}],"predecessor-version":[{"id":18126,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16448\/revisions\/18126"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/16447"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=16448"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=16448"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=16448"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}