How much audio do you need? Use a controlled learning curve Plan speech-data volume with a fixed evaluation set, nested training subsets, slice-level errors and a transparent decision about the next collection. There is no universal number of hours that makes a speech model production-ready. The answer depends on the starting model, task, source diversity, reference quality, training procedure and required behavior. A useful budget decision compares controlled additions of relevant data on a fixed evaluation set, then asks whether the remaining errors are best addressed by more examples, better labels or a change in the model. First distinguish adaptation from training from scratch Adapting an existing recognition model and learning a model from a new initialization are different experiments. A model’s previous training affects what the new examples need to contribute. A transcription service, a multilingual recognizer and a single-speaker synthesis project also use different targets, so their hour counts should not be compared as one market category. Published corpus sizes describe particular research resources. For instance, [LibriSpeech](https://www.openslr.org/12) documents approximately 1,000 hours of read speech; its existence does not establish a minimum for another project. Start by naming the model, task and baseline evaluation result rather than importing an hours rule from an unrelated experiment. Define the decision before buying a tranche Write down a measurable failure: incorrect delivery dates in customer calls, missed technical terms in interviews, or poor recognition on a particular recording channel. Decide which evaluation slices represent that failure and how many independent speakers or sessions support them. Keep the evaluation set outside collection and model-selection decisions as far as practical. Also define the operational consequence. A small change in average word error can coexist with unchanged errors on account numbers. A synthesis project may need pronunciation review and listener judgments rather than recognition WER. Choose a measurement that reflects the product and record where it remains an imperfect proxy. Illustrative experiment: three nested subsets The following plan is an experiment design, not reported performance or a recommended universal purchase size. Nested subsets mean the 5-hour set is included in the 10-hour set, which is included in the 20-hour set. The evaluation recordings remain fixed throughout. Run Data change What stays controlled What to inspect Baseline No new task-specific training Evaluation IDs and scoring policy Current error categories and slice counts A 5 reviewed hours Starting checkpoint, recipe and stated compute budget Whether the task can improve under this setup B 10 reviewed hours including A Same evaluation and documented training changes Marginal change and stability across runs C 20 reviewed hours including B Same evaluation; record update counts Whether the extra volume adds coverage or repeats the same sources Coverage check A smaller alternative with new speakers or channels Comparable task definition and budget accounting Whether diversity addresses the observed gap more effectively Keep data, compute and composition from moving invisibly If one experiment trains for more updates, uses cleaner transcripts and adds new domains at the same time, its result cannot be attributed to hours alone. Record each change and either hold it fixed or describe the experiment as a combined intervention. Equal epochs and equal optimizer steps control different things when dataset sizes change. Report the actual composition: accepted speech duration, speaker/session counts, recording conditions and label review level. Twenty hours from one recurring host can offer fewer independent conditions than a shorter collection from new sources. Conversely, broad diversity can be unhelpful if none of it resembles the target task. The evaluation question decides which comparison matters. Read the curve as a diagnostic, not a guarantee Suppose a hypothetical score improves between A and B but changes little between B and C. That observation could indicate insufficiently novel data, inconsistent labels, inadequate optimization, measurement noise or a model limit. It does not prove a universal law of diminishing returns or identify one correct corpus size. Repeat key runs when feasible and report variability instead of presenting one lucky result as a stable gain. Inspect error examples from each important slice and keep their denominators visible. A reduction measured on a very small subgroup may warrant another review sample before a purchasing decision. Count usable speech and review work separately State whether purchased hours mean wall-clock files, speech-bearing regions, distinct source duration or summed speaker time. Overlapping speakers and repeated derivatives can change those totals. Keep rejected files and excluded music or silence visible in a reconciliation report, rather than silently counting them as model-ready material. Annotation effort is another axis. Correcting systematic transcript omissions may be more valuable than adding raw hours with the same defect. Ask for a pilot that tests the annotation policy as well as the acoustic coverage, and make quote units explicit before comparing providers. Choose a stopping or expansion rule Before running the experiment, state the conditions for the next decision. Examples include an agreed error reduction on a business-critical slice, stable performance across independently grouped evaluation sources, or a cost ceiling for accepted examples. These are project criteria, not thresholds supplied by this article. If the target behavior remains weak, identify the next intervention with the strongest evidence: targeted recruitment, new environments, adjudicated labels, a model adjustment or additional analysis. AIPodcast can discuss sourcing against those gaps. There is no fixed hour count, delivery promise or guaranteed accuracy improvement implied by an inquiry. Can ten hours be enough? It can be a useful experiment size for some settings, but adequacy depends on the starting model, coverage, labels and evaluation target. Test the actual setup rather than treating ten hours as a readiness certificate. Should training duration be equal across the runs? Document the choice. Fixed update counts and fixed epochs answer different questions as data grows. Avoid claiming an hours-only effect when training effort also changed. What if more data does not improve the score? Inspect reference quality, distribution, training controls and uncertainty before buying another bulk tranche. A flat score is evidence about the tested setup, not proof that all additional data is useless. Research method: Fiund research desk; public sources, explicit worked-example labels and project-specific availability review. Editorial contact: hello@fiund.com. Sources Transformers automatic speech recognition task guide: https://huggingface.co/docs/transformers/en/tasks/asr LibriSpeech ASR corpus, SLR12: https://www.openslr.org/12